close

Вход

Забыли?

вход по аккаунту

?

el%3A20010779

код для вставкиСкачать
capacitance, feedback resistance, and the gain of the intemal differential amplifier. The bandwidth is not linearly degraded by
large input gates since the expected capacitive load from MSM or
pin photodetectors at the amplifier input will be large (0.5pF) in
contrast to the additional parallel capacitance of large input gates
(0.1 pF). Furthermore, the feed-forward gain of the frst stage
amplifier is directly proportional to the bandwidth of the feedback
transimpedance amplifier [4].
The use of input transistors with higher thresholds than the
PFET load and NFET current source transistors further augment
both the gain and bandwidth of the differential amplifier. Diode
connected PFET current sourcesiloads are constructed from zero
threshold intrinsic transistors in order to limit capacitive loading
on the output node of the amplifier for a given resistive load,
while intrinsically doped NFET current sources also maximise current per unit area. The NFET input transistors are regular, 0.7V
threshold transistors, since they exhibit a higher g, at equal currents to zero threshold intrinsic devices. Higher g,,, input transistors result in a higher gain bandwidth product for this differential
amplifier.
The diode connected PFETs in this design enable the PFET
current sources to automatically bias themselves in a high gain
operating region without the need for complicated biasing circuitry. Thus, the receiver circuit requires only one bias, in addition
to V,, and ground.
PHANG, K., and JOHNS, D.: ‘A CMOS optical preamplifier for
wireless infrared communictions’, IEEE Trans. Circuits Syst., 1999,
46, (7), pp. 852-859
JAYAKUMAR, A., BUSTOS, M., CHESKIS, D., PIETRUCHA, S., BONELLI, M.,
AL-KURAN, s., and SCHEINBERG, N.: ‘3-v MSM TIA for gigabit
ethernet’. J. Solid-state Circuits, 2000, 35, (9)
WOODWARD, T.K.,
KRISHNAMOORTHY, A.V.,
ROZIER, R.G.,
and
LENTINE, A.L.:
‘Low-power, small-footprint gigabit Ethernet-
compatible optical receiver circuit in 0.25 pm CMOS’, Electron.
Lett., 2000, 36, (17), p. 1489
INGELS, h4., and STEYAERT, M.: ‘A 1-gbis, 0.7-pn CMOS optical
receiver with full rail-to-rail output swing’, IEEE J. Solid Srate
Circuits, 1999, 34, (7), p. 971
HEATLEY, D.J.: ‘Optical receivers’ in FRANCA, Y.T.JOSE E. (Ed.):
‘Design of analog-digital VLSI circuits for telecommunications and
signal processing’ (Prentice Hall, New Jersey, 1994), Chap. 5
AIGaN/AIGaN double-heterojunction
ultraviolet light-emittingdiodes grown by
metal organic chemical vapour deposition
M.M. Wong, J.C. Denyszyn, C.J. Collins,
U. Chowdhury, T.G. Zhu, K.S. Kim and R.D. Dupuis
Light emitting diodes using an Al,Ga,,N/AlyGal_yN doubleheterojunction have been demonstrated with an ultraviolet
emission peak at h = 321nm with a linewidth of 7 . 7 ~ 1 .The
devices were grown by metal organic chemical vapour deposition
on c-plane sapphire substrates.The active region is composed of a
single layer of Al,Ga,_,N (x = 0.23) and the cladding layers are ptw and n-type Al,Ga,N 0, = 0.45). The light output from the
diodes was measured through the sapphire substrate in a )-side
down’ configuration. At a current of 35 mA DC, an output power
of -39 nW was measured without any coatings on the device.
Introduction: The quality of 111-nitride semiconductors grown by
1474/$
Fig. 3 Photograph of 45 x 7 0 p receiver and output buffer in SOS
Results: Fig. 2 shows the operation of this circuit at 650Mbit/s
with a differential optical input and at 900MbiUs with a single
optical input. The data were taken with a 0.5 pF MSM bonded to
each input, and a VCSEL source incident upon an MSM. Some
speckle and jitter is due to the VCSELs and MSMs. However, the
receiver works well in this real situation with both single and differential inputs. Gain is reduced when a single input is used. The
eye diagrams demonstrate that the receiver will work well at both
low and high data rates. Additional jitter may complicate detection of a clock signal for high data rates although output signals
are still detectable up to 600MHz. While the output of the
receiver is rail-to-rail internally, the output buffers will not drive a
50 Q load rail-to-rail, thus the output levels on the eye diagram are
low. At these data rates, we measured a power consumption of
5mW for the full receiver circuit at 3.3V. Fig. 3 shows a 45 x
70 preceiver and output buffer in SOS.
0 IEE 2001
2 July 2001
Electronics Letters Online No: 200I0808
DOI: 10.1049/el:20010808
A. Apse1 and A.G. Andreou (Department of Electrical and Computer
Engineering, Johns Hopkins University, 3400 North Charles Street,
Baltimore, MD, USA)
References
1
GUPTA, M.C.:
‘Handbook of photonics’ (CRC Press, Boca Raton,
USA, 1997)
2
’
INGELS, M., VAN DER PLAS, J.c.G., and STEYAERT, M.: ‘A CMOS
18 THz-ohm 240 Mb/s transimpedance amplifier and 155 Mb/s leddriver for low cost optical fiber links’, IEEE J. Solid State Circuits,
1994, 29, (12), p. 1552
1188
metal organic chemical vapour deposition (MOCVD) has continued to improve, as demonstrated by the recent success in
improved optoelectronic devices such as ultra-high-brightness
light-emitting diodes [ 11 and room-temperature continuous wave
(CW) injection lasers [I - 31. There is a growing interest in compact ultraviolet (UV) light sources, with possible applications such
as chemical sensors, lighting, and optical storage. Many structures
have been investigated, including quantum-well diodes [4, 51 and
double-heterostructure diodes [6, 71. In this Letter we report an
AlGaN/AIGaN double-heterostructure back-emitting light-emitting diode (LED) with an emission peak centred at h = 321 nm.
We believe this to be the shortest wavelength reported for a semiconductor light-emitting diode.
(x = 0.49/y = 0.23) LED
structures of this work are grown by low-pressure metal organic
chemical vapour deposition (MOCVD) in an EMCORE TurboDisc D125 UTM high-speed rotating-disk reactor on 2 inch diameter c-plane (0001) sapphire substrates. The AlGaN epitaxial
layers are gown at pressures -50Torr and the GaN contact layer
is grown at -2OOTorr in a hydrogen ambient using adduct-purified (TMGa), trimethylaluminium (TMAI), and ammonia (NH,).
The n-type and p-type dopants employed are silane (Si&) and
bis(cyclopentadieny1)magnesium (Cp,Mg), respectively. A twotemperature growth process is employed with a low-temperature
AIN buffer layer gown at -540°C and the high-temperature (HT)
device layers grown at -1060°C. First, a 700nm-thick lightly
doped n-Ab,60Ga,,,40N:Si
‘window’ layer is grown, followed by a
‘transition’ layer with an alloy composition graded from x = 0.60
to x = 0.49. This layer is designed to reduce the ‘spikes’ at the
conduction and valence bands at the heterojunction interface. A
200nm-thick n+ &,49Ga,,,51N:Si n-type contact layer is then
grown. The p-side of the device structure consists of a 15Onm
active layer, a 200 nm p +
unintentionally-doped %,,,Ga,,,,N
&,4pGa,,,5,N:Mg p-type cladding layer, and is finally completed
with another graded composition layer to the 25nm-thick p+
GaNMg contact layer. Growth conditions were optimised for
layer thickness uniformity and good quality of the high-AI-compoFubricution: The AlxGal,N/AlyGal,N
ELECTRONICS LETTERS
13th September2001
Vol. 37
No. 19
sition epitaxial layers. To determine the presence of any strain
effects, an X-ray diffraction reciprocal space map was also generated for this wafer. From these measurements, the undoped
AlGaN (x = 0.23) active region is fully strained relative to the
thicker AlGaN (x = 0.49) p and n cladding regions.
The fabrication of the LEDs employs standard processing steps
that we developed for GaN and AlGaN pin photodetectors [8 - 111.
First, the Mg acceptor atoms are activated using rapid thermal
annealing at 850°C for 10min. The diode mesas are then defined
by etching the AlGaN layers through window openings patterned
in a l00nm-thick Si02 mask using reactive ion etching (RIE)
employing BCI, and SiC14. After removing the etch mask and
cleaning the wafer, an SiOz layer is deposited to passivate the
devices. Through openings defined in the passivation layer, Ni/Au
and TUAI contacts are evaporated onto the p-type GaN and ntype &,49Gq,51Nlayers, respectively. The wafer is then annealed
at 750°C for 2min to alloy the ohmic contacts.
I
I
I
p+ GaN:Mg
contact
(25”
p+AIGaN:Mg
grading
(1Onm)
I
A, nm
p+AIGaN:Mg
contact
(200nm)
AIGaN:ud
active
(150nm)
n+AIGaN:Si
contact
(200nm)
n-AIGaN:Si
grading
(45”
n-AIGaN:Si
window
(700nm)
buffer layer
1203/31
Fig. 3 Electroluminescence spectrum (300 K ) from device in Fig. 2
The electroluminescence spectrum, taken at Z = 35mA DC
(65.4A/cm2), from a typical LED, is shown in Fig. 3. The EL
emission from these diodes at il = 321 nm corresponds to approximately 23% AI composition in the active region. This agrees well
with the alloy compositions determined from 0-28 X-ray diffraction. The UV output power at 35mA DC was measured to be
-39nW using a calibrated W-enhanced Si photodiode and a
close-coupling geometry to attempt to collect as much light as possible from one side of the diode. Several diodes have produced
output powers in this range. Owing to the high series resistance,
higher DC drive currents produce heating effects, which resulted
in a slowly increasing light output beyond this current.
I
(25nm)
sapphire substrate
1203Jil
Fig. 1 Schematic diagram of layer structure for AlGaN/AlGaN LED
grown by MOCVD on (0001) sapphire substrate
-
10‘’
diodes emitting at shorter wavelengths are desirable for many
applications. Fig. 1 is a schematic diagram of the AlGaN LED
structure we have employed. The detailed forward I-V characteristics are shown in Fig. 2. Fitting of these data to the ideal diode
equation yields an ideality factor of n = 7.8 in the low current
regime and a series resistance of R, = 745Q in the higher current
regime, indicating that further improvement of our diodes is necessary for optimum performance. In particular, the use of higherconductivity AlGaN cladding layers is necessary.
Conclusion: We have demonstrated short-wavelength W-emitting
AlGaN/AlGaN double-heterojunction light-emitting diodes grown
by metal organic chemical vapour deposition. The diodes are
designed to be mounted p-side down and to emit through the
transparent sapphire substrate to provide a better heatsink for
high-power operation. The peak emission wavelength is h =
3211x11 (3.862eV), with a linewidth of 7.7nm. We believe that
these are the shortest-wavelengthLEDs yet reported.
Acknowledgments: This work was partially sponsored by the
Office of Naval Research under contracts N00014-95-1-1302 and
N00014-99-1-0479 (J.C. Zolper), as well as the National Science
Foundation under grants ESC-0080409 and BES-0086736, and the
DARPA-sponsored Center for Bio-OptoelectronicSensor Systems
under contract MDA9720010020 (D.A. Honey).
70
I 0-3
10-5
0 IEE 2001
18 June 2001
Electronics Letters Online No: 20010779
D 01: IO. 1049/el:20010779
a
5 10-7
i
M.M. Wone. J.C. Denvszvn. C.J. Collins. U. Chowdhurv. T.G. Zhu.
K.S. Kim 2nd R.D. Duphs (Microelectronics Research- Center, The
5
University of Texas at Austin, PRC/MER 1.606D-R9900, Austin, T X
78712-1100, USA)
I 0-9
E-mail: dupuis@mail.utexas.edu
10-1’
References
1 NAKAMURA, s., and FASOL, G.: ‘The blue laser diode’ (Springer-
10-13
voltage, V
m
2
Verlag, Berlin, 1997)
AKASAKI, I., SOTA, s., SAKAI, H.,
TANAKA,T., KOIKE, M.,
and
AMANO, H.: ‘Shortest wavelength semiconductor laser diode’,
Electron. Lett., 1996, 32, (12), pp. 1105-1106
3
BULMAN, G.E.,
U
Fig. 2 I- V characteristics of typical 2 4 0 diameter
~
device
Results: Previously, workers have reported III-N diodes with emission wavelengths as short as il = 333 nm (3.723eV) [5]. However,
no output power was given for these diodes. As discussed above,
ELECTRONICS LETTERS
13th September 2001
Vol. 37
DOVERSPIKE, K.,
SHEPPARD, S.T.,
WEEKS, T.W.,
KONG, H.S.,
DIERINGER, H.M.,
EDMOND, J.A.,
BROWN, J.D.,
SWINDELL, J.T., and SCHETZINA, J.F.: ‘Pulsed operation lasing in a
cleaved-facet InGaN/GaN MQW SCH laser grown on 6H-SiC’,
Electron. Lett., 1997, 33, (18), pp. 1556-1557
No. I9
1189
4
HAN, J . ,
CRAWFORD, M.H.,
SHUL, R.J.,
FIGIEL, J.J.,
BANAS, M.,
ZHANG, L., SONG, Y K., ZHOU, H., and NURMIKKO, A.V.: ‘AIGaNIGaN
quantum well ultraviolet light emitting diodes’, Appl. Phys. Lett.,
1998, 73, pp. 1688-1690
5 KINOSHITA, A., HIRAYAMA, H., AINOYA, M., AOYAGI, Y., and
HIRATA, A.:
‘Room-temperature operation at 333 nm of
Al,,,,Ga, ,J7N/AIn
&iao,5N quantum-well light-emitting diodes with
Mg-doped superlattice layers’, Appl. Phys. Lett., 2000, 77, pp. 175-
sound suppression parameter decision system. The transformation
of the input noisy speech, s(n), to the frequency spectrum, G(k), is
performed using fast Fourier transform (FFT). Update of the
average long-term spectrum magnitudes, INnlmodM(k)l,
for the mth
frame is carried out as follows:
177
6
7
and AMANO, H : ‘Room temperature ultraviolet/blue
light emitting devices based on AlGaNiGaN multi-layered
structure’. Extended abstracts of 1992 Int. Conf. Solid State
Devices and Materials, Tsukuba, Japan, 1992, pp. 327-329
AKASAKI, I.,
OTSUKA, N., TSUJIMURA, A., HASEGAWA, Y., SUGAHARA, G., KUME, M.,
and BAN. Y . : ‘Room temperature 339 nm emission from
AI, ,3Ga0.87N/A1,,
,,,Ga,,,,N double heterostructure light-emitting
diode on sapphire substrate’, Jpn. J. Appl. Phys., 2000, 39, pp.
L445-L448
8
EITING, C.J.,
SHELTON, B.s.,
GRUDOWSKI, P.A.,
PARK, J . S ,
LAMBERT, D.J.H.,
and DUPUIS, R.D.: ‘Characteristics of Mg-doped CaN
grown by metalorganic chemical vapor deposition’, J. Electrochem.
Soc., 1997, 144, pp. L219-L221
9 EITING. c J., GRUDOWSKI, P.A., and DUPUIS, R.D.: ‘P- and N-type
doping of GaN and AlCaN epitaxial layers grown by metalorganic
chemical vapor deposition’, J. Electron. Mat., 1998, 27, (4), pp.
206-209
I O CARRANO, J.c., LI, T., GRUDOWSKI, P.A., EKING, c J., DUPUIS, R.D., and
CAMPBELL,J.C.:
‘Comprehensive characterization of metalsemiconductor-metal ultraviolet photodetectors fabricated on
single-crystal GaN, J. Appl. Phys., 1998, 83, pp. 6148-6160
11
where G,,, = max,lG(k)l, K (= 512) is the FFT sequence length,
and M (= 600), i.e. 6s, corresponds to the blowing period of Bhe
siren sound. The parameter a is set to zero if m < 600 and is set to
0.5 otherwise. In accordance with the periodicity of siren sounds,
the siren sound component in the noisy speech is emphasised by
which is used in the siren sound funcalculating the INmmodM(k)l,
damental frequency detection mentioned below.
LAMBERT, D.J.H., WONG, M.M., CHOWDHURY, U., COLLINS, C . , LI, T.,
KWON, H.K ,
SHELTON, B.s.,
ZHU, T.G ,
CAMPBELL, J c . ,
and
DUPUIS, R.D.: ‘Back illuminated AlGaN solar-blind photodetectors’,
Appl. Phjls. Lett., 2000, 77, (12), pp. 1900-1902
Background noise suppressor for private
mobile radio
S. Sasaki and T. Fumoto
A background noise suppression system used for preprocessing a
speech codec for private mobile radio is described. The system is
able to suppress the sounds of fire engine sirens and other
background noises. Employing the system enabled the sound
articulation of synthetic speech to be improved more than 5% at a
signal-to-noise ratio of 5dB.
I
(593/21
Fig. 2 Block diagram of siren sound suppression parameter decision system
The siren sound fundamental frequency, F,, is extracted as the
frequency f a t which the sum of the rth harmonic magnitudes of
spectrum Ar becomes maximum. Ar is given by
(2)
where fs is the sampling frequency, int(x) is the nearest integer to
x, and L is the number of summed harmonic magnitudes of the
spectrum. The search range forfis from 187.5 to 1OOOHz at intervals of 15.625Hz.
0
coded
speech
.I
noise
suppression
using SS
siren sound
I
1
2
3
time, s
4
5
L59313(
Fig. 3 Typical example pattern of variations in fundamental frequency
of siren sounds
In the detection of siren sounds, the first step is to calculate the
mean squared error (MSE) between the time series of F, and the
prememorised typical pattem of variation in siren sound fundamental frequency. A typical example pattem is shown in Fig. 3.
The siren flag is then set to ON if the MSE is less than the threshold value and is set to OFF otherwise. To avoid abrupt time fluctuations,
is compensated using the typical pattern and the
compensated fundamental frequency F ; is output as the result.
The noise suppressor standardised in TIAEIMS-127 [2] is used
as the noise suppression system using the SS shown in Fig. 1, the
role of which is to suppress stationary noise such as vehicular and
babble noise. In the proposed NS, the system also suppresses the
F ; and its harmonic components of the input noisy speech when
the siren flag indicates ON.
$iJ-
I
7
I
speech
I
I
\ noise
suppressed
speech
I
I
I
I
I
I
I
data
L _ _ _ _ _ _ _ _ _ _ _ _ _ - - - - _ _ _ J- -
Fig. 1 System block diagram
Structure of noise suppression system: As shown in Fig. 1, the proposed noise suppressor (NS) consists of a siren sound suppression
parameter decision system and a noise suppression system using
the spectral subtraction (SS) method. The noise suppressor frame
size is lOms (80 samples). Fig. 2 is a block diagram of the siren
‘I
I90
ll
typical patterns of
variations in fundamental
frequency of siren sounds
Inntrodziction: The use of a speech codec in private mobile radio
(PMR) is important so as to maintain intelligibility and clarity in
the case of channel errors and when there is acoustic background
noise. However, when PMR systems are used in the area of firefighting or police activity, the siren sounds that are acoustically
added to the input speech can considerably degrade the quality of
the synthetic speech. In this Letter, we describe a background
noise suppression system that can suppress the siren sounds of
Japanese fire engines as well as stationary noises such as vehicular
noise. We also report the results of subjective performance tests
for the noise suppression system used for preprocessing a 1.6kbit/s
speech codec [l] developed for PMR.
-
Subjective test results: Intelligibility testing under background
noise conditions was performed using a sound articulation test. In
this test, participants listened to and wrote down 100 Japanese syllables that were arranged randomly and processed by the NS and
speech coders. Sound articulation as expressed in the test results is
the percentage of correct answers with a sound (a consonant or a
vowel) as a unit. Sound articulation of more than 80% indicates
speech quality is acceptable in general communication. Four sets
of 100 Japanese syllables spoken by two male and two female
speakers were used for this testing. The acoustic background noise
ELECTRONICS LETTERS
13th September 2001
Vol. 37
No. I9
Документ
Категория
Без категории
Просмотров
0
Размер файла
418 Кб
Теги
3a20010779
1/--страниц
Пожаловаться на содержимое документа