28.3. Kaushik Roy Dept. of ECE, Purdue University W. Lafayette, IN 47907, U. S. A.
|
|
- Lilian Fletcher
- 5 years ago
- Views:
Transcription
1 8.3 Novel Sizig Algorithm for Yield Improvemet uder Process Variatio i Naometer Techology Seug Hoo Choi Itel Corporatio Hillsboro, OR 974, U. S. A. seug.h.choi@itel.com Bipul C. Paul Dept. of ECE, Purdue Uiversity W. Lafayette, IN 4797, U. S. A. paulb@ec.purdue.edu Kaushik Roy Dept. of ECE, Purdue Uiversity W. Lafayette, IN 4797, U. S. A. kaushik@ec.purdue.edu ABSTRACT Due to process parameter variatios, a large variability i circuit delay occurs i scaled techologies affectig the yield. I this paper, we propose a sizig algorithm to esure the speed of a circuit uder process variatio with a certai degree of cofidece while maitaiig the area ad power budget withi a limit. This algorithm estimates the variatio i circuit delay usig statistical timig aalysis cosiderig both iterad itra-die process variatio ad resizes the circuit to achieve a desired yield. Experimetal results o several bechmark circuits show that oe ca achieve up to 9% savigs i area (power) usig our algorithm compared to the worst-case desig. Categories ad Subject Descriptors B.8.[Performace ad Reliability]:Performace Aalysis ad Desig Aids Geeral Terms: Algorithms, Performace, Desig, Reliability.. INTRODUCTION As silico idustry is movig towards the ed of the roadmap, the device parameters (such as chael legth, oxide thickess, threshold voltage, radom placemet of dopats i chael, etc) are expected to have large variatios. Cosequetly, a large variability i performace amog differet chips is expected. The process variatios ca be classified as systematic or radom. While systematic variatios are determiistic i ature ad are caused by the structure of a particular gate ad its topological eviromet, radom variatios are upredictable i ature. Radom variatios iclude variatios i the effective chael legth of devices, dopig profiles, oxide thickess ad trasistor width. Variatios i dopig profile are very importat i advaced techologies because it may lead to potetially large chage i threshold voltage []. Furthermore, itrisic fluctuatios are idepedet of trasistor locatio o a chip. The process parameter fluctuatios caot be elimiated by exteral cotrol of the maufacturig process ad hece, a statistical desig methodology is required cosiderig the radomess of the process parameter variatio. Various aspects of the process parameter variatio icludig methodology, aalysis, sythesis ad modelig are addressed i []. Covetioal sizig tools size the gates to optimize area ad power cosumptio while meetig the desired delay costrait [3], [4]. Usually these tools fid the critical poits of the circuit through static timig aalysis, which affect the critical path delay. The tool the sizes the trasistor widths to meet the desired delay costrait while keepig the This research was supported by DARPA MARCO Gigascale Silico Research Ceter (SA373JB), SRC (78.) ad Itel Corporatio. Permissio to make digital or hard copies of all or part of this work for persoal or classroom use is grated without fee provided that copies are ot made or distributed for profit or commercial advatage ad that copies bear this otice ad the full citatio o the first page. To copy otherwise, or republish, to post o servers or to redistribute to lists, requires prior specific permissio ad/or a fee. DAC 4, Jue 7-, 4, Sa Diego, CA, USA. Copyright 4 ACM /4/6 $5.. power cosumptio ad area withi a limit. However, due to radom process parameter variatio, a large umber of chips may ot meet the required delay. Cosider a example pdf (probability desity fuctio) of delay show i Figure due to process variatio. I this example, the distributio is assumed to be ormal [5]. This figure shows that 5% of the total umber of dies will ot meet the desired delay costrait, which will affect the fial yield drastically. Oe way to couter this effect is to set the target delay cosiderig worst case process variatio. For example, oe ca choose the 6σ poit i Figure as the target delay for desigig uder worst case process variatio. Cosequetly, while the yield is expected to improve sigificatly, the area ad power overhead to meet the worst case delay costrait may ot be acceptable. This is because, oly a very few dies will have the worst case delay due to process variatio, ad settig the target delay based o those dies will result i uacceptable power cosumptio i most of the dies. Hece, beyod a certai poit the improvemet i yield will be masked by the icrease i the area ad the power overhead. Furthermore, resizig the gate also chages the delay spectrum of the circuit (i.e. σ also chages alog with the mea of delay distributio). This is because the variatio i the trasistor threshold voltage (hece, the variatio i delay) is a strog fuctio of trasistor width due to the radom placemet of dopats i very short-chael devices [6]. A proper desig techique is therefore, ecessary to achieve optimum yield with miimum icrease i the area ad power overhead. Several attempts have bee made to model the effect of process variatio o delay usig statistical timig aalysis [7],[8],[9]. These aalyses cosidered iter- ad itra-die process variatio ad i some cases also modeled spatial correlatios betwee trasistors [8]. However, attempts have rarely bee made to size the gates cosiderig the statistical ature of process variatio. I [], the gate sizig cosiderig a statistical delay model was proposed usig o-liear programmig. However, the use of o-liear programmig makes it less practical i real circuits. I this paper, we propose a statistical desig techique cosiderig both iter- ad itra-die variatio of process parameters. The idea is to resize the trasistor widths with miimal icrease i area ad power cosumptio while improvig the cofidece that the circuit meets the delay costrait uder process variatio. We have developed a sizig Normalized umber of dies Desiged delay Delay (ps) Figure. pdf of delay due to process variatio: a example. 454
2 tool usig Lagragia relaxatio algorithm [] for global optimizatio of trasistor widths. The algorithm first estimates the expected delay variatio at the primary output of a give circuit based o statistical static timig aalysis. Usig the delay distributio obtaied at the primary output ad the desired yield, the algorithm optimizes the area by icreasig the size of trasistors to achieve desired delay while reducig the trasistor sizes i the off critical paths. The cotributio of this work is to provide a sizig algorithm to esure the speed of a circuit uder process variatios with a certai degree of cofidece while keepig the area ad power budget withi a limit. We cosider the variatios ot oly i chael legth, width, oxide thickess, threshold voltage of the trasistor but the effect of radom placemet of dopats as the domiat parameters for process variatio. While the variatios i chael legth, width ad oxide thickess are expected to have spatial correlatio betwee adjacet trasistors [8], radom placemet of dopats make every trasistor i the circuit idepedet. We model these variatios i our aalysis ad icorporate them ito the sizig tool. We also cosider iter- ad itra-die variatio of process parameters to achieve more realistic desig. We used the proposed sizig tool to sythesize several ISCAS bechmark circuits ad compare the estimated yield with the circuit sythesized by covetioal sizig tool. The rest of the paper is orgaized as follows. I sectio, we discuss the statistical timig aalysis of a circuit cosiderig both iter- ad itra-die variatios. Sectio 3 describes the sizig algorithm based o Lagragia relaxatio i detail. I sectio 4, we discuss the experimetal results o ISCAS bechmark circuits. Sectio 5 draws the coclusio.. STATISTICAL TIMING ANALYSIS Statistical static timig aalysis is performed to estimate the variability i circuit delay uder process variatio. Process variatios ca be categorized as iter-die ad itra-die variatios. Due to iter-die variatios, the same device o a chip ca have differet characteristics across differet dies (i.e., dies from oe wafer, from wafer to wafer, ad from wafer lot to wafer lot). Itra-die variatios, o the other had, are the variatios of trasistor characteristics withi a sigle chip. Both iter- ad itra-die variatios are expected to be truly radom i ature i future techologies. While itra-die variatios i terms of trasistor legth, trasistor width ad oxide thickess, are expected to exhibit spatial correlatios amog devices located close to each other, radom placemet of dopats i sub-5 aometer trasistors is expected to make every trasistor i a die idepedet i terms of threshold voltage. We icorporate both iter- ad itra-die variatios i our timig aalysis. The most accurate way of icorporatig the process variatio effects ito timig aalysis is to perform a full-scale trasistor-level Mote- Carlo simulatio of a circuit, which requires large computatioal overhead. Hece, i our aalysis the effect of process parameter variatios o the gate delay is pre-characterized ad accessed o the fly durig statistical timig aalysis. The pre-characterizatio table cotais statistical iformatio o the delay of a gate cosiderig process variatio. It is assumed that iter- ad itra-die process variatios are statistically idepedet [8]. This reduces the complexity of statistical timig aalysis because the effects of iter- ad itra-die process parameter variatio o delay ca be aalyzed i isolatio. For example, we model the variatio i trasistor legth L total as the sum of iter-die variatio (L iter ) ad itra-die variatio ( L itra ) as follows. σ = σ + σ Ltotal Liter Litra where σ is the stadard deviatio. Accordigly, the effect of the variatio i trasistor legth o the delay ca be simplified to [8]: σ delay, total = σ delay, iter + σ delay, itra The effects of variatio i the threshold voltage, oxide thickess, ad trasistor width o delay are also icorporated i a similar way. As explaied above, iter- ad itra-die variatios are aalyzed i isolatio. That is, σ delay,iter ad σ delay,itra are idepedetly calculated ad combied to obtai the overall distributio of delay. Durig statistical timig aalysis, the sigal arrival time is calculated at each gate by propagatig the delay from the primary iput i the circuit. Uder process variatio, the sigal arrival time is also a distributio which is propagated durig the timig aalysis. The variatio i the arrival time is obtaied from the statistical iformatio stored i the precharacterizatio table. We maitai two tables - oe for iter-die variatio ad the other for itra-die variatio. I the followig subsectios, we explai how to propagate the effect of process variatio durig statistical timig aalysis cosiderig both iter- ad itra-die variatio.. Iter-die variatio Cosiderig that trasistor parameters remai costat withi a sigle die, evaluatio of the effect of iter-die variatio o circuit delay is straightforward. For example, due to iter-die variatio, if the device legth i a die becomes 'L+ L', this will remai the same for all trasistors i that die. Hece, uder iter-die process variatios, all idetical gates i a die will have the same delay. It is therefore, easy to pre-characterize the delay of gates for all possible combiatio of process parameters. We geerate the pre-characterizatio table through Mote-Carlo simulatio usig SPICE cosiderig the iter-die variatios i threshold voltage, oxide thickess, trasistor width ad trasistor legth. We assume ormal distributio [5] for all iter-die process parameter variatios. It is also assumed that the correspodig gate delay variatio ca be modeled as ormal distributio. Figure (a) shows oe example of delay variatio for a iverter obtaied usig Normalized umber of dies Normalized umber of dies SPICE results Normal approx Delay (ps) (a) Delay variatio of a gate. Delay variatio.8 Normal approx Delay (ps) (b) Delay variatio of a die Figure. Normal approximatio of iter-die delay variatio 455
3 N(µ S,σ S ) S N(µ A,σ A ) A I I I S S I A A σ Ao = f( σ A, σ A, µ A, µ A, σ S, σ S, S, S ) σ So = g ( σ A, σ A, µ A, µ A, σ S, σ S, S, S ) Figure 3. -iput NAND gate example of calculatig itra-die delay variatio SPICE Mote-Carlo simulatio for BPTM 7m techology [] uder iter-die variatio. It ca be observed from the figure that the above assumptio is reasoable. Each poit i the delay distributio represets the delay of a gate i a die for a correspodig process corer. The circuit delay is calculated for all process corers ad the overall delay distributio for the circuit uder iter-die process variatio is obtaied. Figure (b) shows the delay distributio at the primary output of a example circuit (ISCAS c43) uder iter-die variatio. It ca be see from the figure that the distributio ca also be approximated as ormal distributio.. Itra-die variatio Ulike iter-die variatio, trasistors withi a die are expected to have differet process parameters uder itra-die variatio. I static timig aalysis, sigal arrival time at the output of a gate is calculated by addig the gate delay to the sigal arrival time at the iput. Cosiderig itra-die process variatio, both gate delay ad iput sigal arrival time are to be cosidered as radom variables. For example, i a two-iput gate, the radom variable A o, which represets the worst-case arrival time at the output, ca be expressed as A o = max( A + D, A + D ) () where A ad A are the radom variables which represet iput arrival times, respectively. Radom variables D ad D represet the correspodig pi-to-pi delays. The f Ao (x), the probability desity fuctio for A o, ca be calculated as below. f Ao( x) = f A D ( x) FA D ( x) FA D ( x) f A D ( x) F A (x) represets the probability that A x. Although we assume that A ad D (A ad D ) have ormal distributios, this does ot guaratee that A o ca also be modeled by ormal distributio. However, authors i [3] showed that the error i assumig ormal distributio for A o is egligible. Therefore, i our aalysis, A o is assumed to have ormal distributio. The relatioship show i eq. () imposes additioal costraits o the pre-characterizatio of multiple iput gates uder itra-die variatio. Ulike the switchig uder iter-die variatio (where both iput arrival time ad slope are cosidered to be a sigle value i a die), both iput arrival time ad slope are ow represeted by a distributio. Let us cosider the switchig of a two-iput NAND gate show i Figure 3. The arrival time at the output, A, depeds o the iput sigals, I ad I, ad the gate delay. Iput sigals cosist of sigal arrival time (A, A ) ad the slope (S, S ). I our aalysis, both sigal arrival time ad slope are cosidered as ormal distributios. Therefore, the statistical property (σ Ao, σ So ) of the sigal at the output of the gate is a fuctio of iput arrival times, slopes ad temporal correlatios betwee them. Cosiderig this, we geerate the pre-characterizatio table for σ (itradie variatio), while the mea of the arrival time ad slope at the output S A Out N(µ S,σ S ) N(µ A,σ A ) Out S o A o are calculated o the fly usig Sakurai s delay model [5]. For the ease of aalysis, the temporal proximity of two latest-arrivig iputs is cosidered for multiple iput gates..3 Radom placemet of dopats I scaled CMOS devices, there exists a statistical fluctuatio i the umber of dopats, which ca be traslated ito a threshold-voltage variatio [6]. This discrete dopat effect o threshold voltage variatio is icorporated by employig the followig equatio [6] ito the simulator while geeratig the pre-characterizatio table for both iterad itra-die variatio. σ Vth = q Cox NaWdm 3LW σ Vth represets the stadard deviatio of threshold variatio due to radom placemet of dopats, q is electro charge, C ox is the oxide capacitace, N a is the substrate dopig cocetratio, ad W dm represets the maximum depletio layer width. L ad W are the chael legth ad width of the trasistor, respectively. 3. SIZING ALGORITHM FOR YIELD IMPROVEMENT I this sectio, a gate-sizig algorithm is proposed to improve the yield of a circuit uder process variatio. The algorithm proposed here is based o a well-kow techique to a oliear optimizatio problem: Lagragia relaxatio (LR) []. First, we explai the sizig algorithm based o LR ad the describe our proposed algorithm for sizig cosiderig process variatio. 3. Lagragia relaxatio Che et. al [4] proposed the use of LR for simultaeous sizig of gate ad itercoects of a combiatioal circuit to optimize the total area while maitaiig a delay costrait. The covergece of the algorithm was prove ad the optimality was verified. I our experimets, it is assumed that there are o itercoect compoets i the circuit. However, this algorithm ca be exteded to icorporate the itercoect sizig as well, as explaied i [4]. Figure 4 shows a example circuit represetatio for LR. The circuit cosists of gates, which are to be resized, ad s primary iputs. Logic gates ad primary iputs are called compoets. I additio, we add two virtual compoets, oe coectig all primary iputs (compoet 9 i Figure 4) ad the other coectig primary outputs (compoet ). Therefore, for a circuit with gates ad s iputs, there are +s+ compoets. Edge umbers follow their driver gates, i.e., the output of gate i is deoted as edge i. Compoets ad edges are umbered i reverse topological order. Our objective is to miimize the total area (or equivaletly the power cosumptio) which ca be represeted by Σα i x i,,..., where x i is the gate size ad α i is a arbitrary costat multiplier for gate i, which ca vary depedig o the objective of optimizatio. Covetioally, the gates are sized (i.e., all trasistors i a particular gate are sized by the = 5 (gates), s = 3 (primary iputs) compoets & 9: virtual compoets Figure 4. Circuit represetatio for Lagragia relaxatio 456
4 same factor) to achieve miimum area of the circuit while meetig a give delay costrait. I our aalysis, i additio to the delay costrait at the primary output, A, we also have the yield costrait, γ Hece, the sizig problem is formulated as follows. Miimize αi xi Subject to Di A i p Li xi Ui Yield γ γ p P i =,..., L i ad U i represet the lower boud ad upper boud of the size of gate i, respectively. P is the set of possible paths i a circuit. D i represets the delay of gate i i a path p. Compared to the origial problem formulated i [4], we added a extra costrait for yield. Note that the complexity of the problem is expoetially depedet o the umber of compoets i the circuit (O(e )). To reduce the complexity to a liear oe, the delay costraits o all the paths are trasformed ito the delay costraits o each gate i the circuit. Therefore, the sizig problem (which is called the primal problem; PP) is redefied as follows. PP: Sizig for yield Miimize αixi Subject to a j A j iput() /* outputs * / () a j + D j ai i =,..., j iput() i Di ai i = +,..., + s /* iputs / Li xi Ui i =,..., Yield γ γ a i represets the sigal arrival time at edge i ad D i is the delay associated with gate i. Note that the path-based problem is trasformed to a global problem where A is ow the costrait o the circuit delay, ot o ay specific path i a circuit. I solvig this problem, PP is first traslated ito a mathematical equatio itroducig a Lagragia multiplier λ [] for each costrait o arrival time as follows. Miimize : Lλ ( x, a) = αi xi + λ j ( a j A ) j iput( ) (3) + s + λ ji ( a j + Di ai ) + λmi ( Di ai ) j iput() i + λ ji correspods to the iput edge j ad output edge i of gate i. m i λ mi is equal to +s+ (virtual iput ode). Miimizig L λ usig LR cosists of iteratig the followig two steps: ) calculatig the optimal size of the circuit for the curret λ values ad ) updatig λ to the directio of the optimal solutio. Calculatio of the optimal size ivolves the sizig of each gate i such a way that L λ is 84.% µ A σ A Normal Figure 5 Calculatig A i sizig algorithm Delay costrait A Yield γ Iitialize λ Calculate optimal size x i,opt for λ Perform statistical timig aalysis - Iter-die variatio - Itra-die variatio Calculate A from A ad delay variatio Modify λ based o A Σαx - Q(λ) < error boud? yes optimal sizig Figure 6. Sizig algorithm for yield improvemet locally miimized. Sice D i i L λ is a fuctio of gate size x i, the optimal size of a gate ca be obtaied by solvig dl λ /dx i =. I [4], Elmore delay model was used for D i, however, we use Sakurai's delay model [5] for better accuracy i our aalysis. While updatig λ i step ), arrival-time iformatio at each gate iput/output of a circuit is utilized. That is, λ for the ext iteratio is determied by the curret status of delay costraits imposed o each gate after calculatig the optimal size (step )) [4]. Miimizig L λ provides the miimum size of a circuit while satisfyig the delay costrait, A at the primary output. A more detailed explaatio icludig mathematical proofs o sizig algorithm based o LR ca be foud i [4]. 3. Sizig cosiderig process variatio I this subsectio, we explai the proposed algorithm for resizig cosiderig process variatio. Covetioal sizig method based o LR algorithm explaied above cosiders the delay costrait A ad also the circuit delay as costat values. However, i our aalysis delay at the primary output is represeted by a probability desity fuctio (pdf) cosiderig the process variatio. The is obtaied usig statistical timig aalysis as explaied i sectio. Also ote that ulike covetioal sizig method, we have itroduced a additioal costrait for yield as show i eq. (). We icorporate this by modifyig the delay costrait based o the at the primary output. Figure 5 shows a example o how the yield costrait is itroduced ito the sizig algorithm. Assumig a ormal distributio of delay at the primary output, the modified delay costrait, A should be equal to A - σ i order to achieve, for example, 84.% yield uder process variatio, where σ is the stadard deviatio of the pdf. Similarly, for ay differet type of delay distributio ad yield costrait, the delay costrait ca be modified accordigly. Figure 6 shows the flow diagram of the proposed sizig algorithm with yield cosideratio. It starts with a give delay costrait A ad yield costrait γ. The the iitial λ values are chose so that λ i Ω λ, where Ω λ represets the set of λ values that satisfies the optimality coditio [4]: λ ki = λ jk for k + s i output( k ) j iput( k ) o 457
5 The algorithm the calculates the optimal size of gates for λ ad subsequetly statistical timig aalysis is performed to obtai the delay pdf at the primary output. Note that the chages after each iteratio because the variatio i the threshold voltage (hece, the variatio i delay) is a strog fuctio of trasistor width [6]. Based o the give yield costrait γ ad the, A is modified to a ew costrait A ' as illustrated above. Ulike covetioal sizig methodology, the delay costrait is modified after each iteratio durig the miimizatio of L λ i eq. (3). λ is the updated based o the delay costrait A ' ad the timig iformatio i the circuit. This is repeated util L λ is miimized, i.e., Σαx Q(λ) is less tha a user-defied error boud, where Q(λ) is the optimal solutio for L λ at each iteratio. The modificatio of delay costrait A ' after each iteratio as explaied i our algorithm is however, ot straightforward i a circuit with multiple primary outputs. For example, let us assume that there are two primary outputs i a circuit; gate i ad gate j. Also assume that statistical timig aalysis after k th iteratio shows that µ i < µ j, where µ i ad µ j represet the mea delays at the output of the gate i ad gate j, respectively. Uder the assumptio of ormal ad the target yield of 84.%, there ca be two cases depedig o the value of σ (it is also assumed that the delay distributios at differet primary output gates are correlated):. µ i + σ i < µ j + σ j (Figure 7 (a)) New delay costrait A ' is equal to A - σ j based o the delay variatio at the output of gate j.. µ i + σ i > µ j + σ j (Figure 7 (b)) I this case, the calculatio of A ' should be differet. Note that although gate i is cosidered to produce the worst-case delay uder process variatio, covetioally gate j provides the worst delay without cosiderig the process variatio (µ i < µ j ). Hece, the ew delay costrait, A ' for (k+) th iteratio is equal to A - σ i + (µ j - µ i ). O the other had, if we modify the delay costrait as explaied i case ), the algorithm will size the gates to meet the delay costrait A - σ i at the output of gate j resultig i larger area. It is mathematically prove i [4] that sizig algorithm usig LR always coverges to a optimal solutio. Cosiderig process variatio, we modify the delay costrait i each iteratio based o the delay distributio at the primary output (A ' = A - σ). Hece, the covergece with A ' is guarateed as log as the chage i σ is small. It is observed that the chage i σ due to the chage i circuit size from iteratio to iteratio is cosiderably small, which esures the covergece of the algorithm. I the followig sectio, we discuss the experimetal results o several ISCAS bechmark circuits implemeted usig our proposed sizig algorithm. 4. EXPERIMENTAL RESULTS Our proposed sizig algorithm was used to resize several ISCAS bechmark circuits cosiderig the process parameter variatios. All the circuits were sythesized with BPTM 7m techology []. We assumed 5% (3σ) variatio i all the process parameters such as the width, legth ad the oxide thickess i our aalysis for both iter- ad itra-die variatios. The variatio i the trasistor threshold voltage was govered by the effect of the radom placemet of dopats as metioed i sectio.3. Cosiderig that our sizig algorithm resizes the circuit to meet a certai delay costrait, it is importat to kow the possible rage of the delay that ca be achieved for a circuit ad the desired delay costrait. For example, cosider the area-vs.-delay curve (Figure 8) for ISCAS bechmark circuit c43' obtaied usig LR sizig algorithm. The area σ i µ i σ j A A (a) µ i + σ i < µ j + σ j at gate i at gate j µ j -µ i (the sum of trasistor width) i the plot represets the miimum circuit size for the correspodig delay. I the plot, mi delay represets the miimum delay that ca be achieved by resizig the circuit ad max delay is the circuit delay with all gates havig miimum size. The differece betwee the mi delay ad the max delay is deoted as slack del. Also show i the figure is the stadard deviatio (σ) for the delay variatio due to iter- ad itra-die variatio, which is obtaied from statistical timig aalysis. Let us first assume that our delay costrait correspods to the 5% of slack del. The by usig the proposed algorithm, the circuit ca be resized to meet, for example, 84.% yield for the miimum icrease i the area ( area ). Compared to this area, the icrease i area i the case of delay costrait equal to 9% of slack del is much smaller as show i the figure. Therefore, the effectiveess of our sizig algorithm i terms of miimum icrease i area is depedet o target delay costrait. The differeces i area for the bechmark circuits are summarized i Table. The secod colum shows the slack del of differet circuits i terms of σ (whe the circuit delay is miimum, i.e., mi delay). It varies from.99 to 6.59 for differet circuits. The third ad fourth colums represet the percetage icrease i area for 84.% yield with delay costraits equal to 5% ad 9% of slack del, respectively. As expected, i the case of smaller target delay (5% of slack del ), the icrease i area is larger for the same yield improvemet uder process variatio. We also compared our algorithm with the worst-case desig methodology. For this purpose, we assume the target delay as the 9% of slack del for all bechmark circuits. The compariso results are show i Table. I our experimet, circuits are first sized without cosiderig process parameter variatio. This correspods to the sizig of the circuit for desiged delay show i Figure. The area for the Nomial' desig (without the variability take ito accout, i.e., 5% yield) is show i the third colum of the table which represets the sum of trasistor widths. The fourth colum shows the area after resizig the circuits usig our algorithm cosiderig process variatio. It ca be see that with small icrease i area the yield ca be improved to 84.% (correspodig to σ) uder process variatio. The sixth colum shows σ i σ j A A at gate i at gate j (b) µ i + σ i > µ j + σ j Figure 7. Calculatio of A for differet scearios Total trasistor width (um) mi delay area slack del 5% slack del 9% slackdel σ max delay Circuit delay (ps) Figure 8. Circuit size vs. delay for c43 σ 458
6 Table. The depedece of area o delay costrait slack del / σ area (%) whe target delay is.5 slack del.9 slack del c c c c c c c c c c74l the optimum area for 99.9% yield usig our algorithm. We assume that the target delay is reduced by 3σ (99.9% yield) from the omial desig while resizig uder process variatio. I this case, while the yield improves, the icrease i area is larger tha the previous case. Hece, oe ca make a trade off betwee the yield ad the area (power) budget. Furthermore, i some circuits (c98, c688, c748), it is ot possible to achieve 99.9% yield by resizig the gate. This is because for ay circuit, there is a miimum delay that ca be achieved by resizig the gate. The eighth colum i the table (labeled as Worst desig area') shows the circuit area for the worst-case desig. For worst-case desig, the circuit is sized assumig the worst process corer, i.e., all trasistors will have worst parameter variatios. For example, the trasistor legth is assumed to be 'L + L', where L represets the worst-case variatio i L. Other process parameters are also cosidered i a similar way. It should be oted that while the yield uder process variatio is expected to improve i the worst-case desig, the icrease i area is large. It ca be see that the savig i area is as large as 9% (c74l85) usig our proposed sizig algorithm compared to the worst-case desig. Furthermore, i may cases (deoted as * i the table), it is impossible to size the circuit to achieve the desired delay usig the worst-case desig methodology. Colums ad show the ratio of σ to mea delay with 84.% yield for iter- ad itra-die variatio, respectively. It is observed while the ratio remais almost same for iter-die variatio, the ratio for itra-die variatio varies depedig o the depth of the circuit. It should also be oted that the overall variatio i delay is domiated by iter-die variatio. The last colum shows the rutime of our sizig algorithm for 84.% yield. As explaied i sectio, both iter- ad itra-die process variatios are cosidered through statistical timig aalysis. It is observed i the experimets that the majority of the program rutime is attributed to the aalysis of iter-die variatio. I each iteratio, the circuit was simulated for, differet process corers to icorporate the iter-die variatio as explaied i sectio. The rutime ca be reduced by decreasig the umber of this simulatio while maitaiig the accuracy by usig itelliget samplig techiques as explaied i [6]. 5. SUMMARY We proposed a algorithm to size a circuit for statistical desig cosiderig both iter- ad itra-die variatio. This algorithm estimates the variatio i circuit delay based o statistical timig aalysis ad sizes the circuit to achieve a desired yield with miimum icrease i the area ad power cosumptio. Experimetal results o several bechmark circuits show that the savigs i area (hece the power) ca be as large as 9% usig our algorithm tha the worst-case desig. It was also show that it is ot possible to achieve the desired delay i may circuits usig the worst-case desig methodology. REFERENCES [] X. Tag, V. De, ad J. D. Meidl, Itrisic MOSFET parameter fluctuatios due to radom dopat placemet, IEEE Tras. VLSI Systems, pp , 997. [] C. Visweswariah, Death, taxes ad failig chips, Proc. DAC, pp , 3. [3] J. P. fishbur ad A. E. Dulop, TILOS: A posyomial programmig approach to trasistor sizig, IEEE Tras. CAD, pp , 985. [4] S. Sapatekar, V. B. Rao, P. M. Vaidya, ad S. M. Kag, A exact solutio of the trasistor sizig problem for CMOS circuits usig covex optimizatio, IEEE Tras. CAD, pp , 993. [5] A. Papoulis, Probability, Radom Variables, ad Stochastic Processes, McGraw-Hill, 3 rd editio, 99. [6] Y. Taur ad T. H. Nig, Fudametals of Moder VLSI Devices, Cambridge Uiversity Press, 998. [7] H. F. Jyu, S. Malik, S. Devadas, ad K. W. Keutzer, Statistical timig aalysis of combiatioal logic circuits, IEEE Tras. VLSI Systems, pp. 6-37, 993. [8] A. Agarwal, D. Blaauw, V. Zolotov, S. Sudareswara, M. Zhao, K. Gala, ad R. Pada, Path-based statistical timig aalysis cosiderig iter- ad itra-die correlatios, TAU,. [9] A. Agarwal, D. Blaauw, V. Zolotov, S. Vrudhula, Computatio ad refiemet of statistical bouds o circuit delay, Proc. DAC, pp , 3 [] E. T. A. F. Jacobs, M. R. C. M. Berkelaar, Gate sizig usig a statistical delay model, Proc. DATE, pp. 7-3,. [] M. S. Bazaraa, H. D. Sherali, ad C. M. Shetty, Noliear Programmig: Theory ad Algorithms, Wiley, d editio, 993. [] BPTM, [3] M. R. C. M. Berkelaar, Statistical delay calculatio, a liear time method, TAU, 997. [4] C. P. Che, C.C.N.Chu, ad D.F.Wog, Fast ad exact simultaeous gate ad wire sizig by Lagragia relaxatio, IEEE Tras. CAD, pp.4-5, 999. [5] T. Sakurai ad R. Newto, Delay aalysis of series-coected MOSFET circuits, IEEE JSSC, pp. -3, 99. [6] R. Y. Rubistei, Simulatio ad Mote Carlo method, Joh Wiley ad Sos, 98. Table. Experimetal results of applyig the sizig algorithm to ISCAS bechmark circuits No. of Normal (5% yield) 84.% yield % icrease 99.9% yield % icrease Worst desig % icrease σ iter /µ delay σ itra /µ delay Ru time TR Area (um) area (um) i area area (um) i area area (um) i area (%) (%) (sec) c c * * c * * * * c * * c * * c * * * * c * * * * c c c74l (* sizig failed for the correspodig scheme) 459
A SELECTIVE POINTER FORWARDING STRATEGY FOR LOCATION TRACKING IN PERSONAL COMMUNICATION SYSTEMS
A SELETIVE POINTE FOWADING STATEGY FO LOATION TAKING IN PESONAL OUNIATION SYSTES Seo G. hag ad hae Y. Lee Departmet of Idustrial Egieerig, KAIST 373-, Kusug-Dog, Taejo, Korea, 305-70 cylee@heuristic.kaist.ac.kr
More informationX-Bar and S-Squared Charts
STATGRAPHICS Rev. 7/4/009 X-Bar ad S-Squared Charts Summary The X-Bar ad S-Squared Charts procedure creates cotrol charts for a sigle umeric variable where the data have bee collected i subgroups. It creates
More informationCHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER
95 CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER 5.1 GENERAL Ru-legth codig is a lossless image compressio techique, which produces modest compressio ratios. Oe way of icreasig the compressio ratio of a ru-legth
More informationApplication of Improved Genetic Algorithm to Two-side Assembly Line Balancing
206 3 rd Iteratioal Coferece o Mechaical, Idustrial, ad Maufacturig Egieerig (MIME 206) ISBN: 978--60595-33-7 Applicatio of Improved Geetic Algorithm to Two-side Assembly Lie Balacig Ximi Zhag, Qia Wag,
More informationAPPLICATION NOTE UNDERSTANDING EFFECTIVE BITS
APPLICATION NOTE AN95091 INTRODUCTION UNDERSTANDING EFFECTIVE BITS Toy Girard, Sigatec, Desig ad Applicatios Egieer Oe criteria ofte used to evaluate a Aalog to Digital Coverter (ADC) or data acquisitio
More informationA New Space-Repetition Code Based on One Bit Feedback Compared to Alamouti Space-Time Code
Proceedigs of the 4th WSEAS It. Coferece o Electromagetics, Wireless ad Optical Commuicatios, Veice, Italy, November 0-, 006 107 A New Space-Repetitio Code Based o Oe Bit Feedback Compared to Alamouti
More information(2) The MOSFET. Review of. Learning Outcome. (Metal-Oxide-Semiconductor Field Effect Transistor) 2.0) Field Effect Transistor (FET)
EEEB73 Electroics Aalysis & esig II () Review of The MOSFET (Metal-Oxide-Semicoductor Field Effect Trasistor) Referece: Neame, Chapter 3 ad Chapter 4 Learig Outcome Able to describe ad use the followig:
More informationDesign of FPGA- Based SPWM Single Phase Full-Bridge Inverter
Desig of FPGA- Based SPWM Sigle Phase Full-Bridge Iverter Afarulrazi Abu Bakar 1, *,Md Zarafi Ahmad 1 ad Farrah Salwai Abdullah 1 1 Faculty of Electrical ad Electroic Egieerig, UTHM *Email:afarul@uthm.edu.my
More informationLogarithms APPENDIX IV. 265 Appendix
APPENDIX IV Logarithms Sometimes, a umerical expressio may ivolve multiplicatio, divisio or ratioal powers of large umbers. For such calculatios, logarithms are very useful. They help us i makig difficult
More informationSurvey of Low Power Techniques for ROMs
Survey of Low Power Techiques for ROMs Edwi de Agel Crystal Semicoductor Corporatio P.O Box 17847 Austi, TX 78744 Earl E. Swartzlader, Jr. Departmet of Electrical ad Computer Egieerig Uiversity of Texas
More informationSuper J-MOS Low Power Loss Superjunction MOSFETs
Low Power Loss Superjuctio MOSFETs Takahiro Tamura Mutsumi Sawada Takayuki Shimato ABSTRACT Fuji Electric has developed superjuctio MOSFETs with a optimized surface desig that delivers lower switchig.
More informationDesign of FPGA Based SPWM Single Phase Inverter
Proceedigs of MUCEET2009 Malaysia Techical Uiversities Coferece o Egieerig ad Techology Jue 20-22, 2009, MS Garde,Kuata, Pahag, Malaysia MUCEET2009 Desig of FPGA Based SPWM Sigle Phase Iverter Afarulrazi
More informationAC : USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM
AC 007-7: USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM Josue Njock-Libii, Idiaa Uiversity-Purdue Uiversity-Fort Waye Josué Njock Libii is Associate Professor
More informationELEC 350 Electronics I Fall 2014
ELEC 350 Electroics I Fall 04 Fial Exam Geeral Iformatio Rough breakdow of topic coverage: 0-5% JT fudametals ad regios of operatio 0-40% MOSFET fudametals biasig ad small-sigal modelig 0-5% iodes (p-juctio
More informationINCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION
XIX IMEKO World Cogress Fudametal ad Applied Metrology September 6, 9, Lisbo, Portugal INCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION Dalibor
More informationWAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI
WAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI Muhammad Kabir McGill Uiversity Departmet of Electrical ad Computer Egieerig Motreal, QC H3A 2A7 Email: muhammad.kabir@mail.mcgill.ca Carlos Christofferse
More informationICM7213. One Second/One Minute Timebase Generator. Features. Description. Ordering Information. Pinout. August 1997
August 997 Features Guarateed V Operatio Very Low Curret Cosumptio (Typ).... µa at V All Outputs TTL Compatible O Chip Oscillator Feedback Resistor Oscillator Requires Oly Exteral compoets: Fixed Capacitor,
More informationx y z HD(x, y) + HD(y, z) HD(x, z)
Massachusetts Istitute of Techology Departmet of Electrical Egieerig ad Computer Sciece 6.02 Solutios to Chapter 5 Updated: February 16, 2012 Please sed iformatio about errors or omissios to hari; questios
More informationAnalysis of SDR GNSS Using MATLAB
Iteratioal Joural of Computer Techology ad Electroics Egieerig (IJCTEE) Volume 5, Issue 3, Jue 2015 Aalysis of SDR GNSS Usig MATLAB Abstract This paper explais a software defied radio global avigatio satellite
More informationSummary of pn-junction (Lec )
Lecture #12 OUTLNE iode aalysis ad applicatios cotiued The MOSFET The MOSFET as a cotrolled resistor Pich-off ad curret saturatio Chael-legth modulatio Velocity saturatio i a short-chael MOSFET Readig
More informationCompound Controller for DC Motor Servo System Based on Inner-Loop Extended State Observer
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 6, No 5 Special Issue o Applicatio of Advaced Computig ad Simulatio i Iformatio Systems Sofia 06 Prit ISSN: 3-970; Olie ISSN:
More informationSensors & Transducers 2015 by IFSA Publishing, S. L.
Sesors & Trasducers 215 by IFSA Publishig, S. L. http://www.sesorsportal.com Uiversal Sesors ad Trasducers Iterface for Mobile Devices: Metrological Characteristics * Sergey Y. YURISH ad Javier CAÑETE
More informationHigh Speed Area Efficient Modulo 2 1
High Speed Area Efficiet Modulo 2 1 1-Soali Sigh (PG Scholar VLSI, RKDF Ist Bhopal M.P) 2- Mr. Maish Trivedi (HOD EC Departmet, RKDF Ist Bhopal M.P) Adder Abstract Modular adder is oe of the key compoets
More informationx 1 + x x n n = x 1 x 2 + x x n n = x 2 x 3 + x x n n = x 3 x 5 + x x n = x n
Sectio 6 7A Samplig Distributio of the Sample Meas To Create a Samplig Distributio of the Sample Meas take every possible sample of size from the distributio of x values ad the fid the mea of each sample
More informationTehrani N Journal of Scientific and Engineering Research, 2018, 5(7):1-7
Available olie www.jsaer.com, 2018, 5(7):1-7 Research Article ISSN: 2394-2630 CODEN(USA): JSERBR 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
More informationDIGITALLY TUNED SINUSOIDAL OSCILLATOR USING MULTIPLE- OUTPUT CURRENT OPERATIONAL AMPLIFIER FOR APPLICATIONS IN HIGH STABLE ACOUSTICAL GENERATORS
Molecular ad Quatum Acoustics vol. 7, (6) 95 DGTALL TUNED SNUSODAL OSCLLATOR USNG MULTPLE- OUTPUT CURRENT OPERATONAL AMPLFER FOR APPLCATONS N HGH STABLE ACOUSTCAL GENERATORS Lesław TOPÓR-KAMŃSK Faculty
More informationProblem of calculating time delay between pulse arrivals
America Joural of Egieerig Research (AJER) 5 America Joural of Egieerig Research (AJER) e-issn: 3-847 p-issn : 3-936 Volume-4, Issue-4, pp-3-4 www.ajer.org Research Paper Problem of calculatig time delay
More informationE X P E R I M E N T 13
E X P E R I M E N T 13 Stadig Waves o a Strig Produced by the Physics Staff at Colli College Copyright Colli College Physics Departmet. All Rights Reserved. Uiversity Physics, Exp 13: Stadig Waves o a
More informationAnalysis and Optimization Design of Snubber Cricuit for Isolated DC-DC Converters in DC Power Grid
Aalysis ad Optimizatio Desig of Subber Cricuit for Isolated DC-DC Coverters i DC Power Grid Koji Orikawa Nagaoka Uiversity of Techology Nagaoka, Japa orikawa@st.agaokaut.ac.jp Ju-ichi Itoh Nagaoka Uiversity
More informationUnit 5: Estimating with Confidence
Uit 5: Estimatig with Cofidece Sectio 8.2 The Practice of Statistics, 4 th editio For AP* STARNES, YATES, MOORE Uit 5 Estimatig with Cofidece 8.1 8.2 8.3 Cofidece Itervals: The Basics Estimatig a Populatio
More informationOutline. Motivation. Analog Functional Testing in Mixed-Signal Systems. Motivation and Background. Built-In Self-Test Architecture
Aalog Fuctioal Testig i Mixed-Sigal s Jie Qi Dept. of Electrical & Computer Egieerig Aubur Uiversity Co-Advisors: Charles Stroud ad Foster Dai Outlie Motivatio ad Backgroud Built-I Self-Test Architecture
More informationA New Design of Log-Periodic Dipole Array (LPDA) Antenna
Joural of Commuicatio Egieerig, Vol., No., Ja.-Jue 0 67 A New Desig of Log-Periodic Dipole Array (LPDA) Atea Javad Ghalibafa, Seyed Mohammad Hashemi, ad Seyed Hassa Sedighy Departmet of Electrical Egieerig,
More informationGeneral Model :Algorithms in the Real World. Applications. Block Codes
Geeral Model 5-853:Algorithms i the Real World Error Correctig Codes I Overview Hammig Codes Liear Codes 5-853 Page message (m) coder codeword (c) oisy chael decoder codeword (c ) message or error Errors
More informationAn Optimal Test Pattern Selection Method to Improve the Defect Coverage
A Optimal Test Patter Selectio Method to Improve the Defect Coverage Yuxi Tia, Michael R. Grimaila, Weipig Shi ad M. Ray Mercer Departmet of Electrical Egieerig Texas A&M Uiversity, College Statio, Texas
More informationPROJECT #2 GENERIC ROBOT SIMULATOR
Uiversity of Missouri-Columbia Departmet of Electrical ad Computer Egieerig ECE 7330 Itroductio to Mechatroics ad Robotic Visio Fall, 2010 PROJECT #2 GENERIC ROBOT SIMULATOR Luis Alberto Rivera Estrada
More informationPHY-MAC dialogue with Multi-Packet Reception
PHY-AC dialogue with ulti-packet Receptio arc Realp 1 ad Aa I. Pérez-Neira 1 CTTC-Cetre Tecològic de Telecomuicacios de Cataluya Edifici Nexus C/Gra Capità, - 0803-Barceloa (Cataluya-Spai) marc.realp@cttc.es
More informationRadar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1
Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,
More informationCross-Layer Performance of a Distributed Real-Time MAC Protocol Supporting Variable Bit Rate Multiclass Services in WPANs
Cross-Layer Performace of a Distributed Real-Time MAC Protocol Supportig Variable Bit Rate Multiclass Services i WPANs David Tug Chog Wog, Jo W. Ma, ad ee Chaig Chua 3 Istitute for Ifocomm Research, Heg
More informationTest Time Minimization for Hybrid BIST with Test Pattern Broadcasting
Test Time Miimizatio for Hybrid BIST with Test Patter Broadcastig Raimud Ubar, Maksim Jeihhi Departmet of Computer Egieerig Talli Techical Uiversity EE-126 18 Talli, Estoia {raiub, maksim}@pld.ttu.ee Gert
More informationA study on the efficient compression algorithm of the voice/data integrated multiplexer
A study o the efficiet compressio algorithm of the voice/data itegrated multiplexer Gyou-Yo CHO' ad Dog-Ho CHO' * Dept. of Computer Egieerig. KyiigHee Uiv. Kiheugup Yogiku Kyuggido, KOREA 449-71 PHONE
More informationEfficient Feedback-Based Scheduling Policies for Chunked Network Codes over Networks with Loss and Delay
Efficiet Feedback-Based Schedulig Policies for Chuked Network Codes over Networks with Loss ad Delay Aoosheh Heidarzadeh ad Amir H. Baihashemi Departmet of Systems ad Computer Egieerig, Carleto Uiversity,
More informationMeasurement of Equivalent Input Distortion AN 20
Measuremet of Equivalet Iput Distortio AN 2 Applicatio Note to the R&D SYSTEM Traditioal measuremets of harmoic distortio performed o loudspeakers reveal ot oly the symptoms of the oliearities but also
More informationR. W. Erickson. Department of Electrical, Computer, and Energy Engineering University of Colorado, Boulder
R. W. Erickso Departmet of Electrical, Computer, ad Eergy Egieerig Uiversity of Colorado, Boulder 4.2.2. The Power MOSFET Gate Source Gate legths approachig oe micro p - p Cosists of may small ehacemetmode
More informationAME50461 SERIES EMI FILTER HYBRID-HIGH RELIABILITY
PD-94595A AME5046 SERIES EMI FILTER HYBRID-HIGH RELIABILITY Descriptio The AME Series of EMI filters have bee desiged to provide full compliace with the iput lie reflected ripple curret requiremet specified
More informationCOMPRESSION OF TRANSMULTIPLEXED ACOUSTIC SIGNALS
COMPRESSION OF TRANSMULTIPLEXED ACOUSTIC SIGNALS Mariusz Ziółko, Przemysław Sypka ad Bartosz Ziółko Departmet of Electroics, AGH Uiversity of Sciece ad Techology, al. Mickiewicza 3, 3-59 Kraków, Polad,
More informationMaximum Voltage Variation in the Power Distribution Network of VLSI Circuits with RLC Models Λ
Due to the type 3 fots used, please icrease the magificatio to view Maximum Voltage Variatio i the Power Distributio Network of VLSI Circuits with RLC Models Λ Sudhakar Bobba Su Microsystems Ic. 91 Sa
More informationSingle Bit DACs in a Nutshell. Part I DAC Basics
Sigle Bit DACs i a Nutshell Part I DAC Basics By Dave Va Ess, Pricipal Applicatio Egieer, Cypress Semicoductor May embedded applicatios require geeratig aalog outputs uder digital cotrol. It may be a DC
More information4. INTERSYMBOL INTERFERENCE
DATA COMMUNICATIONS 59 4. INTERSYMBOL INTERFERENCE 4.1 OBJECT The effects of restricted badwidth i basebad data trasmissio will be studied. Measuremets relative to itersymbol iterferece, usig the eye patter
More informationAdaptive Resource Allocation in Multiuser OFDM Systems
Adaptive Resource Allocatio i Multiuser OFDM Systems Fial Report Multidimesioal Digital Sigal Processig Malik Meherali Saleh The Uiversity of Texas at Austi malikmsaleh@mail.utexas.edu Sprig 005 Abstract
More informationCombined Scheme for Fast PN Code Acquisition
13 th Iteratioal Coferece o AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT- 13, May 6 8, 009, E-Mail: asat@mtc.edu.eg Military Techical College, Kobry Elkobbah, Cairo, Egypt Tel : +(0) 4059 4036138, Fax:
More informationThe Institute of Chartered Accountants of Sri Lanka
The Istitute of Chartered Accoutats of Sri Laka Postgraduate Diploma i Busiess ad Fiace Quatitative Techiques for Busiess Hadout 02:Presetatio ad Aalysis of data Presetatio of Data The Stem ad Leaf Display
More informationHybrid BIST Optimization for Core-based Systems with Test Pattern Broadcasting
Hybrid BIST Optimizatio for Core-based Systems with Test Patter Broadcastig Raimud Ubar, Masim Jeihhi Departmet of Computer Egieerig Talli Techical Uiversity, Estoia {raiub, masim}@pld.ttu.ee Gert Jerva,
More informationAn Adaptive Image Denoising Method based on Thresholding
A Adaptive Image Deoisig Method based o Thresholdig HARI OM AND MANTOSH BISWAS Departmet of Computer Sciece & Egieerig Idia School of Mies, Dhabad Jharkad-86004 INDIA {hariom4idia, matoshb}@gmail.com Abstract
More informationOn Parity based Divide and Conquer Recursive Functions
O Parity based Divide ad Coquer Recursive Fuctios Sug-Hyu Cha Abstract The parity based divide ad coquer recursio trees are itroduced where the sizes of the tree do ot grow mootoically as grows. These
More informationRoberto s Notes on Infinite Series Chapter 1: Series Section 2. Infinite series
Roberto s Notes o Ifiite Series Chapter : Series Sectio Ifiite series What you eed to ow already: What sequeces are. Basic termiology ad otatio for sequeces. What you ca lear here: What a ifiite series
More informationA Heuristic Method: Differential Evolution for Harmonic Reduction in Multilevel Inverter System
Iteratioal Joural of Computer ad Electrical Egieerig, Vol. 5, o. 5, October 013 A Heuristic Method: Differetial Evolutio for Harmoic Reductio i Multilevel Iverter System P. Jamua ad C. Christober Asir
More informationSEVEN-LEVEL THREE PHASE CASCADED H-BRIDGE INVERTER WITH A SINGLE DC SOURCE
SEVEN-LEVEL THREE PHASE CASCADED H-BRIDGE INVERTER WITH A SINGLE DC SOURCE T. Porselvi 1 ad Ragaath Muthu 1 Sri Sairam Egieerig College, Cheai, Idia SSN College of Egieerig, Cheai, Idia E-Mail: tporselvi@yahoo.com
More informationA New Basic Unit for Cascaded Multilevel Inverters with the Capability of Reducing the Number of Switches
Joural of Power Electroics, ol, o, pp 67-677, July 67 JPE --6 http://dxdoiorg/6/jpe67 I(Prit: 98-9 / I(Olie: 9-78 A ew Basic Uit for Cascaded Multi Iverters with the Capability of Reducig the umber of
More informationDelta- Sigma Modulator with Signal Dependant Feedback Gain
Delta- Sigma Modulator with Sigal Depedat Feedback Gai K.Diwakar #1 ad V.Vioth Kumar *2 # Departmet of Electroics ad Commuicatio Egieerig * Departmet of Electroics ad Istrumetatio Egieerig Vel Tech Uiversity,Cheai,
More informationPermutation Enumeration
RMT 2012 Power Roud Rubric February 18, 2012 Permutatio Eumeratio 1 (a List all permutatios of {1, 2, 3} (b Give a expressio for the umber of permutatios of {1, 2, 3,, } i terms of Compute the umber for
More informationReduction of Harmonic in a Multilevel Inverter Using Optimized Selective Harmonic Elimination Approach
ISSN (Olie) : 2319-8753 ISSN (Prit) : 2347-6710 Iteratioal Joural of Iovative Research i Sciece, Egieerig ad Techology Volume 3, Special Issue 3, March 2014 2014 Iteratioal Coferece o Iovatios i Egieerig
More informationLETTER A Novel Adaptive Channel Estimation Scheme for DS-CDMA
1274 LETTER A Novel Adaptive Chael Estimatio Scheme for DS-CDMA Che HE a), Member ad Xiao-xiag LI, Nomember SUMMARY This paper proposes a adaptive chael estimatio scheme, which uses differet movig average
More informationHigh-Order CCII-Based Mixed-Mode Universal Filter
High-Order CCII-Based Mixed-Mode Uiversal Filter Che-Nog Lee Departmet of Computer ad Commuicatio Egieerig, Taipei Chegshih Uiversity of Sciece ad Techology, Taipei, Taiwa, R. O. C. Abstract This paper
More informationInternational Power, Electronics and Materials Engineering Conference (IPEMEC 2015)
Iteratioal Power, Electroics ad Materials Egieerig Coferece (IPEMEC 205) etwork Mode based o Multi-commuicatio Mechaism Fa Yibi, Liu Zhifeg, Zhag Sheg, Li Yig Departmet of Military Fiace, Military Ecoomy
More informationInformation-Theoretic Analysis of an Energy Harvesting Communication System
Iformatio-Theoretic Aalysis of a Eergy Harvestig Commuicatio System Omur Ozel Seur Ulukus Departmet of Electrical ad Computer Egieerig Uiversity of Marylad, College Park, MD 074 omur@umd.edu ulukus@umd.edu
More informationCHAPTER 8 JOINT PAPR REDUCTION AND ICI CANCELLATION IN OFDM SYSTEMS
CHAPTER 8 JOIT PAPR REDUCTIO AD ICI CACELLATIO I OFDM SYSTEMS Itercarrier Iterferece (ICI) is aother major issue i implemetig a OFDM system. As discussed i chapter 3, the OFDM subcarriers are arrowbad
More informationLecture 28: MOSFET as an Amplifier. Small-Signal Equivalent Circuit Models.
hites, EE 320 ecture 28 Page 1 of 7 ecture 28: MOSFET as a Amplifier. Small-Sigal Equivalet Circuit Models. As with the BJT, we ca use MOSFETs as AC small-sigal amplifiers. A example is the so-called coceptual
More informationThe Firing Dispersion of Bullet Test Sample Analysis
Iteratioal Joural of Materials, Mechaics ad Maufacturig, Vol., No., Ma 5 The Firig Dispersio of Bullet Test Sample Aalsis Youliag Xu, Jubi Zhag, Li Ma, ad Yoghai Sha Udisputed, this approach does reduce
More informationLaboratory Exercise 3: Dynamic System Response Laboratory Handout AME 250: Fundamentals of Measurements and Data Analysis
Laboratory Exercise 3: Dyamic System Respose Laboratory Hadout AME 50: Fudametals of Measuremets ad Data Aalysis Prepared by: Matthew Beigto Date exercises to be performed: Deliverables: Part I 1) Usig
More informationA Novel Small Signal Power Line Quality Measurement System
IMTC 3 - Istrumetatio ad Measuremet Techology Coferece Vail, CO, USA, - May 3 A ovel Small Sigal Power Lie Quality Measuremet System Paul B. Crilly, Erik Leadro Boaldi, Levy Ely de Lacarda de Oliveira,
More informationCHAPTER 6 IMPLEMENTATION OF DIGITAL FIR FILTER
CHAPTER 6 IMPLEMENTATION OF DIGITAL FIR FILTER 6.1 INTRODUCTION The digital FIR filters are commo compoets i may digital sigal processig (DSP) systems. There are various applicatios like high speed/low
More informationNovel pseudo random number generation using variant logic framework
Edith Cowa Uiversity Research Olie Iteratioal Cyber Resiliece coferece Cofereces, Symposia ad Campus Evets 011 Novel pseudo radom umber geeratio usig variat logic framework Jeffrey Zheg Yua Uiversity,
More informationThe Detection of Abrupt Changes in Fatigue Data by Using Cumulative Sum (CUSUM) Method
Proceedigs of the th WSEAS Iteratioal Coferece o APPLIED ad THEORETICAL MECHANICS (MECHANICS '8) The Detectio of Abrupt Chages i Fatigue Data by Usig Cumulative Sum (CUSUM) Method Z. M. NOPIAH, M.N.BAHARIN,
More informationNovel Modeling Techniques for RTL Power Estimation
Novel Modelig Techiques for RTL Power Estimatio Michael Eierma Walter Stechele Istitute for Itegrated Circuits Istitute for Itegrated Circuits Techical Uiversity of Muich Techical Uiversity of Muich Arcisstr.
More informationA Bipolar Cockcroft-Walton Voltage Multiplier for Gas Lasers
America Joural of Applied cieces 4 (10): 79-799, 007 N 1546-99 007 ciece Publicatios orrespodig Author: A Bipolar ockcroft-walto Voltage Multiplier for Gas Lasers hahid qbal ad Rosli Besar Faculty of Egieerig
More informationTotal Harmonics Distortion Reduction Using Adaptive, Weiner, and Kalman Filters
Wester Michiga Uiversity ScholarWorks at WMU Master's Theses Graduate College 6-2016 Total Harmoics Distortio Reductio Usig Adaptive, Weier, ad Kalma Filters Liqaa Alhafadhi Wester Michiga Uiversity, liquaa.alhafadhi@yahoo.com
More informationSummary of Random Variable Concepts April 19, 2000
Summary of Radom Variable Cocepts April 9, 2000 his is a list of importat cocepts we have covered, rather tha a review that derives or explais them. he first ad primary viewpoit: A radom process is a idexed
More informationIntroduction to Wireless Communication Systems ECE 476/ECE 501C/CS 513 Winter 2003
troductio to Wireless Commuicatio ystems ECE 476/ECE 501C/C 513 Witer 2003 eview for Exam #1 March 4, 2003 Exam Details Must follow seatig chart - Posted 30 miutes before exam. Cheatig will be treated
More informationMeasurements of the Communications Environment in Medium Voltage Power Distribution Lines for Wide-Band Power Line Communications
Measuremets of the Commuicatios viromet i Medium Voltage Power Distributio Lies for Wide-Bad Power Lie Commuicatios Jae-Jo Lee *,Seug-Ji Choi *,Hui-Myoug Oh *, Wo-Tae Lee *, Kwa-Ho Kim * ad Dae-Youg Lee
More informationNew MEGA POWER DUAL IGBT Module with Advanced 1200V CSTBT Chip
New MEGA POWER DUAL IGBT Module with Advaced 1200V CSTBT Chip Juji Yamada*, Yoshiharu Yu*, Joh F. Dolo**, Eric R. Motto** * Power Device Divisio, Mitsubishi Electric Corporatio, Fukuoka, Japa ** Powerex
More informationJoint Power Allocation and Beamforming for Cooperative Networks
It. J. Commuicatios, etwork ad System Scieces,, 4, 447-45 doi:.436/ijcs..4753 Published Olie July (http://www.scirp.org/joural/ijcs) Joit Power Allocatio ad Beamformig for Cooperative etworks Sodes Maadi,,
More informationPower-Aware Modulo Scheduling for High-Performance VLIW Processors
Power-Aware Modulo Schedulig for High-Performace VLIW Processors Ha-Saem Yu School of Computer Sciece ad Egieerig Seoul Natioal Uiversity hsyu@davici.su.ac.kr Jihog Kim School of Computer Sciece ad Egieerig
More informationIEEE P Wireless Personal Area Networks
November 6 IEEE P8.5-6-398-3-3c IEEE P8.5 Wireless Persoal Area Networks Project Title IEEE P8.5 Workig Group for Wireless Persoal Area Networks (WPANs) Statistical 6 GHz Idoor Chael Model Usig Circular
More informationOptimal Geolocation Updating for Location Aware Service Provisioning in Wireless Networks
Optimal Geolocatio Updatig for Locatio Aware Service Provisioig i Wireless Networks Siri Tekiay Amer Catovic tekiay@adm.jit.edu axc4466@jit.edu New Jersey Istitute of Techology Uiversity Heights, Newark,
More informationPROCEED: A Pareto Optimization-based Circuit-level Evaluator for Emerging Devices
PROCEED: A Pareto Optimizatio-based Circuit-level Evaluator for Emergig Devices Shaodi Wag, Adrew Pa, Chi O Chui, ad Pueet Gupta Departmet of Electrical Egieerig Uiversity of Califoria, Los Ageles Los
More informationWavelength Band Switching in Multigranular Optical WDM Networks
Wavelegth Bad Switchig i Multigraular Optical WDM Networks Vishal Aad Collaborators X. Cao,, Dr. Y. Xiog ad Dr. C. Qiao LANDER, CSE Departmet, SUNY at Buffalo ~1~ Vishal Aad Outlie λ The Problem with preset
More informationFast Sensor Deployment for Fusion-based Target Detection
Fast Sesor Deploymet for Fusio-based Target Detectio Zhaohui Yua*, Rui Ta*, Guoliag Xig*, Cheyag Lu, Yixi Che *Departmet of Computer Sciece, City Uiversity of Hog Kog Departmet of Computer Sciece ad Egieerig,
More informationData Acquisition System for Electric Vehicle s Driving Motor Test Bench Based on VC++ *
Available olie at www.sciecedirect.com Physics Procedia 33 (0 ) 75 73 0 Iteratioal Coferece o Medical Physics ad Biomedical Egieerig Data Acquisitio System for Electric Vehicle s Drivig Motor Test Bech
More informationMessage Scheduling for the FlexRay Protocol: The Dynamic Segment
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 1 Message Schedulig for the FlexRay Protocol: The Dyamic Segmet Ece Gura Schmidt, Member, IEEE, Klaus Schmidt Abstract The FlexRay commuicatio protocol is expected
More informationEfficient Energy Consumption Scheduling: Towards Effective Load Leveling
eergies Article Efficiet Eergy Cosumptio Schedulig: Towards Effective Load Levelig Yua Hog 1, *, Shegbi Wag 2 ad Ziyue Huag 3 1 Departmet of Iformatio Techology Maagemet, Uiversity at Albay, SUNY, 1400
More informationPerformance and Analysis with Power Quality improvement with Cascaded Multi-Level Inverter Fed BLDC Motor Drive
Iteratioal Joural of Recet Advaces i Egieerig & Techology (IJRAET) Performace ad Aalysis with Power Quality improvemet with Cascaded Multi-Level Iverter Fed BLDC Motor Drive 1 N. Raveedra, 2 V.Madhu Sudha
More informationECE 902. Modeling and Optimization of VLSI Interconnects
ECE 90 Modelig ad Optimizatio of VLSI Itercoects (http://eda.ece.wisc.edu/ece90.html Istructor: Lei He Email: he@ece.wisc.edu Office: EH343 Telephoe: 6-3736 Office hour: TR :30-3pm Course Prerequisites
More informationAN ESTIMATION OF MULTILEVEL INVERTER FED INDUCTION MOTOR DRIVE
9 IJRIC. All rights reserved. IJRIC www.ijric.org E-ISSN: 76-3336 AN ESTIMATION OF MULTILEVEL INVERTER FED INDUCTION MOTOR DRIVE K.RAMANI AND DR.A. KRISHNAN SMIEEE Seior Lecturer i the Departmet of EEE
More informationAfter completing this chapter you will learn
CHAPTER 7 Trasistor Amplifiers Microelectroic Circuits, Seeth Editio Sedra/Smith Copyright 015 by Oxford Uiersity Press After completig this chapter you will lear 1. How to use MOSFET as amplifier. How
More informationA Novel Three Value Logic for Computing Purposes
Iteratioal Joural o Iormatio ad Electroics Egieerig, Vol. 3, No. 4, July 23 A Novel Three Value Logic or Computig Purposes Ali Soltai ad Saeed Mohammadi Abstract The aim o this article is to suggest a
More informationOptimal Arrangement of Buoys Observable by Means of Radar
Optimal Arragemet of Buoys Observable by Meas of Radar TOMASZ PRACZYK Istitute of Naval Weapo ad Computer Sciece Polish Naval Academy Śmidowicza 69, 8-03 Gdyia POLAND t.praczy@amw.gdyia.pl Abstract: -
More informationOn the Delay Performance of In-network Aggregation in Lossy Wireless Sensor Networks
O the Delay Performace of I-etwork Aggregatio i Lossy Wireless Sesor Networks Chaghee Joo, Member, IEEE, ad Ness B. Shroff, Fellow, IEEE Abstract I this paper, we study the implicatio of wireless broadcast
More informationSIDELOBE SUPPRESSION IN OFDM SYSTEMS
SIDELOBE SUPPRESSION IN OFDM SYSTEMS Iva Cosovic Germa Aerospace Ceter (DLR), Ist. of Commuicatios ad Navigatio Oberpfaffehofe, 82234 Wesslig, Germay iva.cosovic@dlr.de Vijayasarathi Jaardhaam Muich Uiversity
More informationMike Li Andy Martwick Gerry Talbot Jan Wilstrup
ITC 24 Trasfer Fuctios For The Referece Clock Jitter I A Serial Lik: Theory Ad Applicatios Mike Li Ady Martwick Gerry Talbot Ja Wilstrup Trasfer Fuctios For The Referece Clock Jitter I A Serial Lik: Theory
More informationAME28461 SERIES EMI FILTER HYBRID-HIGH RELIABILITY
PD-94597A AME28461 SERIES EMI FILTER HYBRID-HIGH RELIABILITY Descriptio The AME Series of EMI filters have bee desiged to provide full compliace with the iput lie reflected ripple curret requiremet specified
More information