2.25MH;r nffsci. 'I'he clcvclo~ctl dexiiw meets ihc IS-95 CUMA device spccilicwtions under hoth 3.0V and 2.4V biws contlitioiis.
CrirrcirrsWts; A low voltage p o i w PHEMI' tbr CDMA applicii- tions IiiiS heen dcvcloped. Thc device h;is it double delta-tioped AIG~IAS~GJIA~/CT:IAS s(ructii rc to providc high drain current dcnsity aiicl transcoiiductaiic. 'l'lie size of l l i e layout of [lie dcvicc was mduccd tn iinpruvc the low tlic vnltagc pcrkmxincc. A1 2.4V bias, Ihc dcvdopcd devicc show? an oulpul powcr or 36dHni with
:I powcr added eificicricy nf bI.ST4 arid tlic gaiil of thc dcvice at maximuin output powur i s 8.47dU. Wlicri icstctl under IS-95 CDMA ruodul;ilioii conditions, llic dcvicc iiwcts ilic CIJMA spw ificiitions a1 huth 3.OV illid 2 . 4 V tlinin I h . A1 2 M B m lincar out- put pnwcr hiq CDMA ;ipplication, the devicc lltis ii powcr iitkled eRlcicticy or 373%) ;it 3.0V hias and 30.2%~ a t 2.4V bias. This Ltt- tcr represenis thc first reporl 011 powcr I'lIEMTs for 2.4V CDMA spplications. The dcvicc slioiild bc iipplicahk 10 the ncxl genera- tion of digital wirelcss coininunic;itioii systems.
coritacterl t i ' -p tliodcs without and with Ar iinpl;intntioii, respx-
livcly. l o r the diode williout A r ir~q~lantation in Fig. la, thc poly- Si/% iiilerkicc was frilly h o k i i i mil coinplctc cpitaxial realignmcnt or Ihc poly-Si with respect to thc Si substrate was obscrvcd. This is cousisteiil with thc result reporid in [(I), However, for Ihc diode with AV impl;uit~tion (Pig. Ih), onIy hnlr of thc poly-Si \vas epi- (axidly rciiligned (region I1 of i l c micrograpli). 'L'lic iissocintctl TEM diffraclioii patterns of i~gions A kind B are
also
showii in k'ig, I b . These pallerris indicatc that region A is polycrystalliiie and region B is a niono-cryshl. 1,'ig. 2 sliows a11 cnlwrgcd TEM micrograph 01" rcyiun A, whetc bubble-likc tlcccts i n llic, .. . ,"* a Si oJim & . . . . ~. .
.;
Sipply-Si/Si interfaco oAm
(I Without A V iiiip1;intntioii b With Ar inip1atit;itiun
1)iffrilclion plittcrns of rcgions A and B arc dso sliowri h i Fig. 3 h
implanktl polycrystal line rcgioii CBII he ohserved. lhcsc hul)hle-
like defects are bclicvcd to iict a s getteriiig cciilrcs for As, R aiid t;
580
ElEC
iltoiiis, relarding their diITiisioii into the poly-WSi interface. 1:i.g. 3rr and h diow sindar TEM micrographs D I the cliotlcs which v"impltriilcd with P iiistcarl
or As.
I:or the diode without Ar implantation (Fig. 31~), Ihc polySilSi inlcrfiiw, was alsc iiilly bru- ken and cntnpletc epitilxial rarlignniciit or l h c poly-Si occurred: but for thc diode wilh AT impktntwiion (Fig. 3 4 only half of thc poly-Si m i l t e r w a s epit.iixially miligned (rcgiun U). For this siiiii- ple, II~WOVLY, i m r e micro-twins iliiin those or Ihc A d o p c d diodeof Fig. lb ~ I F C observed. Tlic assuciatcd TEhI diffraction pattenis ot' regioiis A and B for this diorlc ;irc also included in Pig. 3h. They also iiirlic;itc that reghi A nus polycrystalliiic and rccgion B w;is ii single crysial.
23 22 10 m E
:
21 c- 10 loeo 0 PE
2
3
U2
18 10 17 10 0 0.2 0.4 0.6 depth, wn (1) As (ill Ii .. , . ,
.
, . , , , . witlinnt AI- ilnpl;mtatioti- w i h i i r iinphiitntion, IO%III
1;ig 4 shows the secondary i o n inass spcctrometry (SIMS) pro- files of As ;ind T;, Iespcclivcly, of the j i + - p pcrlynittci's 01' Fig. 10
aiid b. I1 can dsn lie seeii Ilia1 llic AY implaii(cd sili~iple liad higher As arid F pwks at the Rp or l h c implaiited Ar in thc poly-Si tiliii.
a lower F' pcak ;it the poly-SYSi inicrfacc alii1 ti shiillowcr As prn- filc i n the utidcrlying silicon suhsirrik Thc lower As aiid 1: coii- ceirlr;ilion rcwlted
in
ii lowcr level of epitxxid rccaligitineiit or ~ h c poly-% willi rcspcct in ilic Si substrate. A similar SIMS profile orP
and 17 Tor ihc j i l . - p poly-emillcr diodc of Fig. 3 sl~owcd the smileresid t.
Clotrcksviuri: In this Isilcr, i t Iras t w n dcmotistraterl that Ar
implantation can rctard the epit;ixial icaligiiriictit in ;in t i t - p poly- einittei' diotlc by cicating bubble-likc dciects which gcttcr F, As, trnd P. This iwhniquc could bc m c h l for improving thc gait) of tlic poly-cmittcr trausislur in hipolar ICs.
Denoising
by optimal
fuzzy
thresholding
in
wavelet
domain
L.-K.
Shilrka~nd
C,
YII'I'l~c cotrrlructioii ol' a hylirid navclelcl lliieshold is prcscnlcd basctl myinpintic iicw-niiiiiiniix lixcd tIircshold rur iioi4c wpIircssiuii. T11c opiitniilily UT ilre prnpnscd Ihrcshold is dctnoristriiicd by ircovcring ii Garissian c n v d o F sinusoidal sigtinl ciiihiddcrl iii (lie adtlitivc wliilc G;iussian inoisc in goniitl iind ~ p u o r sigilal-to-noisc
CIL I I o mil atios.
nti k i u y cull\binitIloth of :in unbiawti risk wliiiiak iinrl a11
Introrluotirw Tn dcnoisiiig based on the discrclc wavelct 1r;lnsCom
(LlWl'), tlic gcricral appro~icl~ i s to apply ii ~Iii-esl~old to tlic LVBVC- h coemcieiits protlticctl hy tlic h w ; i r t l 1 IWT of thc noisc-cor- rupted signal, thcrccby allowing only significant wavclcl coefficients with their magnilutle values grciilcr tlitiii tlic tlircshold tn he iiscd in tlic inverse DWT ([DW'l~) to rcconstrrict [lie origind signul. Conscquciitly, there exists ii tlileiiiina in selling NI tipprqirialc thrculiold: raising ~ h c rhresliolrl to rcducc the i~oisc contribution iii
sigiial rcconsluictioii can inccciisc: [lie signal distorlion, becaiisc it inlay rcsul( in sinall wavclcl coertlcierits due IO the sigixil bciiig cxcliideti in the I D W T :ilkriii~1iveiy, lowcriiig the tlircsliultl 10
iricreasc the signnl coiitribulioii iii signd rccunsli'uction ciin
iiicrcasc noise interfercncc, bwcaiise it m;iy icsult in small wltvcler cocfticicnts tliic IO noise bcing iiiclutlcrl i i i tlic tI>Wl'. To solve thc dilcmmw, ti special fuzzy iiicnihealiip liinction has t m n prcvioiisly
proposctl such that t l c contrilmtion 01' it wvelet cocficicnt h i sig-
nal reconstriction dcpeiids on tlic rclalive significiincc of i t s inag- iiiliide v;iluc [ 11. Ilowever, thc compii1xtion of tlic opriiiial ti~zzy meinlicrship rtmction requircs kntiwledge oE the signal to lie rccov- crccd. To exteiid fuzzy lti~~eestioldiiig tu rhc recovery 01' niikiiown signals, iliis Lcticr proscnls the coiislructiuii of' ii Icvcl-clcpcndeiii
optiiiial tiizzy incinherstiip tiiiichn by coinbitling Skin's unbi;isctl risk cstiuiate (StJKE') ;inti mi asyiiiptolic tiear-miniiixix lixcd threshold to deteriniric ~ h c bcst vducs for ils cotitrolling piraiiie- ~ c r s . Also presciitcd arc ihe coinputcr simulation rcsiilts lo tleinon- stnttc thc optimality of thc propowl Tuzzy thrcshulrliiiy iitiictioii for denoising.
Op/irrim//irzz)~ 6 / ? V 0 d M J k / k g If s,, is llic Signal t o be rcsloi'eti iind 11;~ ir; the aclditivc xcro meui wliitc Qaiissiaii iioisc with miriiincc a?,
the11 tlic rcceiwrl signid is given by
I ' I L = s,, -1. IU,, '/I, = 0 , I ,
...:
N ~ I . (1)whwt N is the iitiinbcr of sainplcs in Ilia rcceivcd signal. Applying ihc IIWT with 11.1 tlccoiiipositioii slagcs lo I ; , yiclds ;ni iipproxiina-
tinn cocficicnt .sequeticc ;it tlic coalsest rcsolutioii, iil,,,, with / I =
0, 1,
...,
( N 2 9
1 , arid :i sct oC delail coctYicicnt sccpieiices at dif-fkrciit mololioiis, I/,,
,,,,, with
~ I I I = I , 2, ..., M niid I I = 0, I ,..., (AQ'
2'")-I. Sincc tlic DiV"" is ~l lincm opcratinn, rlic DW.1' of tlic rwcivetl signal cm be viewcrl ;is ;I lineiir supcrpusitiou of thc 13WT of s,, with a small Iiumbcr of sigiiiticant cuclkieirrs a d tlic DWT or U:, with ;I largc inwilier of mull coeflicicnts. C:oizsc-quciilly, clcnoising btvoiiies ii protilcm of fumtilahig an ;ipprulwi- ale tlireshnlil to cxcludc sinall detail cocfticieiils (prcsumahly gcncralcd h y noiscl in the IDWT, tlmrchy yielding ;1 rertnrcil vcr-
sinii of thc originid sigiid,
i,,,
A fixcd form tlirestiold with some isyniptotic ncar-iniiiimax projxrlics Tor the detail cocficiciils wl ~ l i c J d i rcsoluhn level is givcti by [21
w1ici.e
cients nt the firs1 rcsoliilioii level as
is llic nuisc variaiice cstirii;itcrl iisiiig [lie detail coetTi-
Altliough this clioicc or thrcsliold v;iluc is simple, i t i s kiiowii as ii
conrcrviitivc m e t h t tcncls Lo zero out i i w c dct;iiI coeflicients 11s thc niimbcr of detail cocfliciciits increascs. At thc liinit, the probir- bility d' n detail cociKcicnl. a t tlic mth resolution lcvel Iiaving i i value greatcr thaii tf"" will tend to zero i i s N approaclm =_
Aii altcriwtivc chuicc for yielding ii lowcr Ihrccsliold v d u c crime
Tiom tlic Stciii iinbiasctl risk eslitiwtc (SURIS) based oil miiiimisa- t i m of thc 1nc;in-sqiwed crrors bc~wccii s,, iiiid c*,,, I)y vxyiiig t h e tilrcsiloiri t
aolll
zel.o ull to~p
t;ll:,"" =
wlicre #{Id,,:,,,
I
5f l
ticiiores the nuiiibcr d'detail cocfficiciils IIBV- ing tiiagiiiturlc values less than t. ChnpilwI with ihc fixed form lhrcshold, the heliaviouc or llie SIJRB tlircshold rends to bc crrxtic and leiids to iiiclriticmow
detail coctficicnts i n sigixil rccoristrudioii t l s the iioisc contribiitioii iiwe:iscs.Coiisequeiitly, tlicic is a problem iii sclccting between the fixed
h i m i tlrreshold and thc SUIXE thrcslioltl. Whilc tlic rorincr ciin be
viewed as the nppcr boiind Ixcausc it Icnds 10 'overkill' thc detail
cnefticicnts, I h c iiittel can 1~ vicwcrl i i s [he lower t~otind bccilllse it tcnds l o 'uinderkill' tlic dclail coclTtieiits. A possiblc coiiiproiiiisc is i o usc ii i w z y incinbcrship rmciion to ;~llow thc detail cocfh-
ciei-lts lying hetwccn i,i:"" rind I ~ ; ; ~ ~ ' . to coiilribuie partially in the Sigrid rccons(riiclion according LO tfieir Iiiagnituclc va[ucs (propor- tionid shrinkage OT tlic dctiiil c o c k i e i i t s ) . One parlicular membcr-
ship runc(ion, wliich wils fouiirl to pcri'orin wcll for signal rccovwy
using the l>Wl', is given by
I
I]Similarly, sincc tlic dctail coefficients w i h v;ilucu gicatcr ~ l i a i i ~/,f"'d