Professional Documents
Culture Documents
A Full-Day Tutorial at VLSI Design & Test Symposium July 23, 2008
Lecture 1: Introduction
Vish ani D. Vishwani D Agrawal, Agra al vagrawal@eng.auburn.edu agra al@eng a b rn ed Foster Dai, daifa01@auburn.edu
Auburn University, Dept. of ECE, Auburn, AL 36849, USA
Abstract
This tutorial discusses design and testing of RF integrated circuits (RFIC) It is suitable for engineers who plan work on RFIC but did not (RFIC). have training in that area, those who work on IC design and wish to sharpen their understanding of modern RFIC design and test methods, and engineering managers. It is an abbreviated version of a onesemester t university i it course. Specific S ifi topics t i i l d semiconductor include i d t technologies for RF circuits used in a wireless communications system; basic characteristics of RF devices linearity, noise figure, gain; RF front-end design g LNA, , mixer; ; frequency q y synthesizer y design g p phase locked loop (PLL), voltage controlled oscillator (VCO); concepts of analog, mixed signal and RF testing and built-in self-test; distortion theory, measurements, test; noise theory, measurements, test; RFIC SOCs and their testing. testing
Objectives
To acquire introductory knowledge about integrated circuits (IC) used in radio frequency (RF) communications systems. To learn basic concept of design of RFIC. To learn basic concepts of RFIC testing.
Outline
Introduction to VLSI devices used in RF communications SOC and SIP Functional components Technologies Design concepts Test concepts Basic RF measurements Distortion characteristics Noise SOC testing and built-in self-test (BIST)
4
References
1. 2. 3 3. 4. 5. M. L. Bushnell and V. D. Agrawal, Essentials of Electronic Testing for Digital, Memory & Mixed-Signal VLSI Circuits, Boston: Springer, 2000. J. Kelly and M. Engelhardt, Advanced Production Testing of RF, SoC, and SiP Devices, Boston: Artech House, 2007. B Razavi, B. Razavi RF Microelectronics, Upper Saddle River, River New Jersey: Prentice Hall PTR, 1998. g , C. Plett and F. Dai, , Integrated g Circuit Design g for High-Speed g p J. Rogers, Frequency Synthesis, Boston: Artech House, 2006. K. B. Schaub and J. Kelly, Production Testing of RF and System-on-a-chip Devices for Wireless Communications, Boston: Artech House, 2004.
Schedule
09:30AM 10:00AM 10:00AM 11:00AM 11:00AM 11:30AM 11:30AM 13:00PM 13:00PM 14:00PM 14:00PM 15:00PM 15:00PM 15:30PM 15:30PM 17:30PM Lecture 1 Lecture 2 Break Lecture 3 Lunch Lectures 4 Break Lectures 5-7 Lecture 8 RF Testing II RF BIST Agrawal Dai
6
Introduction RF Design g I
Agrawal Dai
RF Design II
Dai
RF Testing I
Agrawal
An RF Communications System
Superheterodyne Transceiver
0 0
LNA VGA Phase Splitter
LO
Duplexer r
90
LO
ADC DAC 0
PA
VGA
Phase Splitter
LO
90 DAC
RF
IF
BASEBAND
Duplexer r
90 ADC DAC 0
PA Phase Splitter
LO
90 DAC
RF
BASEBAND
8
Components of an RF System
Radio frequency
Duplexer LNA: Low noise amplifier PA: Power amplifier RF mixer Local oscillator Filter
Mixed-signal
ADC: Analog to digital converter DAC: Digital to analog converter
Digital
Digital g signal g processor p (DSP)
Intermediate frequency
VGA: Variable gain amplifier Modulator M d l t Demodulator Filter
Duplexer
TDD: Time-Division Duplexing
Same Tx and Rx frequency RF switch (PIN or GaAs FET) L Less th than 1dB l loss
Rx
fr ft Rx
fr
Tx TDD command
ft
Tx
10
Technologies:
Bipolar CMOS
Technologies:
GaAs SiGe SiG
12
Tecnologies:
Bipolar MOS
Passive Mixer
V(RF) V(LO)
nFET RL
V(IF)
14
Active Mixer
VDD
15
16
Phase Splitter
Splits input signal into two same frequency outputs that differ in phase by 90 degrees. Used for image rejection.
C Vin R
R Vout_1 V t 2 Vout_ C
17
SOC: System-on-a-Chip
All components of a system are implemented on the same VLSI chip. Requires same technology (usually CMOS) used for all components. Components not implemented on present-day SOC: Antenna Power amplifier (PA)
18
Lecture 2: RF Design I
f o = N f ref
+
ffb
Divider N
fo
N is an integer the minimum step size = fr to get a smaller step size, the reference frequency must be made smaller N must be higher in order to generate the same fo larger phase noise (in-band (in band noise magnified 20logN times by the loop) loop).
Frequency synthesizer design I (PLL), FDAI, 2008 21
Fractional-N Concept
If the loop divisor N is a fractional number, e.g., N=K/F, where K and F are integer numbers the minimum step size = fr /F can achieve small step size without lowering the reference frequency loop divisor N can be small in order to generate the same fo better phase noise (in-band noise magnified 20logN times by the loop). How can we design a fractional divider? Divider is a digital block and its output transits only at the input clock edge we can only generate integer frequency divider!! Dual-modulus divider P/P+1: by toggling between the two integer division ratios, a fractional division ratio can be achieved by time-averaging the divider output. As an example, if the control changes the division ratio between 8 and 9 and the divider divides by 8 for 9 cycles and by 9 for 1 cycle and then the 9, process repeats itself, then the average division ratio will be:
N=
8 9 + 9 1 = 8.1 10
22
Transfer function that controls l loop d dynamics i (LPF) fr R ffb Dual Modulus Divider P/P+1 Carry out bit Cout
CLK
1 z-1
F(s)
Step Size =
Controllable Signal Source (VCO)
fr RF
fo
K
yI +
+ log2F yI-1
Fractional Accumulator
f Cout =
Kf clk F
23
Kf clk = F
+
ffb
F(s) Multi-Modulus Controllable Signal Source (VCO) Divider MMD Modulus control nbit
fo =
fo +
fr R
K I + F
+
+
CLK z-1 Fractional Accumulator 1bit
Integer divisor I
+
+
Cout +
log2F
25
Since the comparison frequency is fr/R = 5.12 MHz, the fractional accumulator size can be chosen as: F = f 1 = 5120kHz = 32
r
R 160kHz
160kHz
which can be implemented using a 5-bit accumulator. The accumulator input, i.e., the fine tune frequency word K, can be programmed from 0 to 10 to cover the 11 channels f from 819.2 819 2 MHz MH to t 820.96 820 96 MH MHz with ith step t size i of f 160 kHz kH (th (the fi first t channel h ld does not t require any fractionality). The integer divisor ratio, i.e., the coarse tune frequency word I, can be determined by the channel frequency. For instance, the first channel frequency is synthesized y as:
fr R 0 f I + = r I = 819 .2 MHz F R
which leads to I = 160. Hence, the loop total divisor is given by N = 160 + K/32, where K = 0, 1, 10.
Frequency synthesizer design I (PLL), FDAI, 2008 27
28
id = KCP K PD ( R o )
vc =
I id
R(s)
Crystal Oscillator
Kphase UP DN PFD
e(s)
I
Loop Filter R C1
v c( s )
Magnitude Response
C2
Gain
o(s)
N Divider VCO
K vco s
Phase Response
CS C2 C1C2 C1 + C2
where CS =
o
vc
1 K VCO N s
VCO ( s)
vc ( s)
K VCO s
Phase
1 RC1 1 RCS
VCO = K VCO vc
29
C2 (about C1/10) adds a high frequency pole to clean up high frequency ripple on the control line. line
Without C2 , Cs = C1C2 =0 C1 + C2
90 135 180
1 RC1 1 RCS
30
natural frequency
n =
K NC1
o (dB) R
or
=5
-5 2 1.414 -15 -20 0.1 1 0.707 0.2 0.4 0.7 1 2
damping constant
3dB = n 1 + 2 + 4 + 4 + 2
2 4 2
R 2
KC1 N
3 dB 1 + 2 n
< 1 .5
3dB 2 n = K N
> 1.5
/ n
7 10 0.5 0.3
32
33
UP
CLK
FF
CLK RST
OUT
vo H
CLK RST
FF
IN OUT
DN
RST
Positive edge-triggered D flip flop with active low flip-flop reset and hidden D=1
34
dead zone
Dead Zone
0
T
-I
2 e
Phase Noise P
Frequency Offset
35
Phase/Frequency Detector
1 fr
Phir
rst_
active low w reset active high lock detect reset_ t
Delay
LD
fV 1
D
rst rst_
PhiV
36
VCC
UP+ UPVref
CPout
DOWN+ DOWNVref
37
VCO
K phase
C2
The 2nd-order filter is the highest g order p passive RC filter that can be built without series resistors between the charge pump and the VCO tune line
F (s ) =
C1C2 Cs = C1 + C2
38
vc C3
C2
-20 dB/dec
Legend: 2nd Order PLL 3rd Order PLL 4th Order PLL
F (s ) =
T1 = R1 C1 T2 = R1 C1 C2 Ct T R C 3 3 3
Ct = C1 + C2 + C3
Comparison of open loop gain and phase in a second, third, and fourth order PLL
39
vout (t ) = Vo cos( LO t + n (t ))
n (t ) = p sin ( m t )
Single sideband (SSB) phase noise power spectral density (PSD) to carrier ratio is defined as the ratio of power in one phase modulation sideband per Hertz bandwidth, at an offset away from the carrier, to the total signal power in units of [dBc/Hz]:
rms (f ) =
180
10
PN DSB ( f ) 10
180 2
10
PNSSB ( f ) 10
[deg/
Hz
2 rms
( f )df
40
LO
Frequency Reference
41
PD+LPF
GLPF kVCO/s
VC O
VCO
LPF
PD
output S0
NTF
%N
+
N
1 1 G = 1 N 1+ G / N 1+ G / N
42
Flicker 1/f noise is caused by trapping in the semiconductor material material. Flicker noise corner fc is an empirical parameter depending on the device size and processing. For CMOS, fc is found to be 3~7 kHz typically yp y and for bipolar p transistors fc is about as 50 kHz. Notice that fc has impact only on close-in noise. QL is the loaded Q of the resonant circuit, ranging from 5~20 for on-chip resonator and 40~80 for off-chip tank. Ps is the average signal power at output of the oscillator active device, and F is oscillator effective noise factor.
Frequency synthesizer design I (PLL), FDAI, 2008 44
3. Crystal/CP Intercept
Div id
er n
ois e
Cr y stal n
CP nois
oise
2. CP/VCO Intercept p
nois e
no ise
45
Crys tal n
oise
To
t al
no i
se
CP i no
PD noise
se
O VC
ise no ise no
se noi
F LP
46
-60 60 -70
-80 -90 -100 -110 -120 -130 -140 -150 -160 0.1 1.0 10 100 1000 10000
Parameter
C1 C2 R
47
References
J. Rogers, C. Plett, and F. Dai, Integrated Circuit Design for High-Speed Frequency Synthesis, Boston: Artech House, 2006. F. Dai and C. Stroud, Analog and Mixed-Signal Test Architectures, Chapter 15 in System-on-Chip Test Architectures: Nanometer Design for Testability, Morgan Kaufmann Publishers, 2007. J. Rogers, F. Dai and C. Plett, Frequency Synthesis for Multi-band Wireless Networks, Chapter 15 in Emerging Wireless Technologies -- From System to Transistors, CRC Press, 2007. B. Razavi, RF Microelectronics, Upper Saddle River, New Jersey: Prentice Hall PTR, 1998. J. Rogers, F. F. Dai, M. S. Cavin, and D. G. Rahn, A Fully Integrated MultiBand SD Fractional-N Frequency Synthesizer for a MIMO WLAN Transceiver RFIC IEEE Journal of Solid-State Circuits, vol RFIC, vol. 40 40, no no. 3 3, pp pp. 678-689 678-689, March March, 2005.
Frequency synthesizer design I (PLL), FDAI, 2008 48
Lecture 3: RF Design II
PN = P0 / N0
fo
noise floor
Where Po is the power in the tone at the frequency of oscillation and No is the noise power spectral density at some specified offset from the carrier. Phase noise i i is usually ll specified ifi d i in dBc/Hz dB /H , meaning i noise i at t1 1-Hz H b bandwidth d idth measured d in decibels with respect to the carrier.
Frequency synthesizer design I (PLL), FDAI, 2008 50
(t )
vout =
2 I pulse e C
1 1 cos t 2 2 LC 4 R C
Amplitude
Oscillation frequency
OSC =
1 1 LC 4 R 2 C 2
In most oscillators,
R >> L / C
OSC =
1 LC
51
Rp
-Rn
The addition of negative resistance to the circuit to overcome losses in a) ) a parallel p resonator or b) ) a series resonator.
Vin(s)
+
52
Barkhausen Criterion
Closed loop gain
1 H 1 (s) H 2 (s) = 0
positive feedback with gain larger than H 1 ( j ) H 2 ( j ) = 1 or equal to 1. Barkhausen criterion, which states that for sustained oscillation at constant amplitude, the gain around the loop is 1 and the phase around the loop is 0 or some multiple of 2. For sustained oscillation at constant amplitude, the pole must be on the j axis. For the openl loop analysis l i
H 1 ( j ) H 2 ( j ) = 1
H 1 ( j ) H 2 ( j ) = 2 n
53
Noise in one sideband in a 1Hz bandwidth: noise power in 1Hz BW at wo+w Ltotal(w) = Carrier power Units: dBc/Hz
Frequency synthesizer design I (PLL), FDAI, 2008
(c)
Buffer
L
G G
Amplifier
Amplifier Amplifier
Resonators with feedback. (a) Colpitts Oscillator. (b) Hartley Oscillator (not suitable for IC). (c) -Gm oscillator. v t growth limited
Frequency synthesizer design I (PLL), FDAI, 2008 55
Waveform of an LC resonator with losses compensated. p The oscillation g grows until a practical constraint limits the amplitude.
vin Z in = i = g m X 1 X 2 + j ( X 1 + X 2 ) iin
gm 1 1 Z in = 2 + + C1C 2 jC1 jC 2
g m X 1 X 2 > RL g m > RL 2 C1C 2
f osc 1 = 2
Load reactance needs to equal j(X1+X2) add an inductor
1 2
Oscillation frequency
1 1 1 + L C C 2 3 1
56
57
Impedance seen by b tank k =(1+C1/C2)2/gm For negligible loading on the tank, C1 > 10C2 needs a large cap.
Frequency synthesizer design I (PLL), FDAI, 2008
58
v2 vi gm1v1 v1
ii =
vi = g m1v 1 g m 2 v 2 re1 + re 2
Zi =
gm >
2 gm
2 Rp
60
VCO Mathematical Model Output frequency of an ideal VCO: wout = wFR + Kvco Vc
t
Q = (w0/2) |d/dw |
Vcc M5 D1 Q3 Q4
RB2
Vcc M6 Q5
Rref
Vcc
CTail
Ibg
iout+
ioutRBB
VBias RBB M1 Cc Cc
Q5
Q6
M2
RB1
D2 D3
Q1
RE
IBais
Q2
RE
62
Nout(s) H1(s)
H2(s)
VOSC = A cos[ 0 t + n (t )]
N out ( s ) H 1(s) = N in ( s ) 1 H ( s )
H ( j ) H ( j 0 ) +
dH d
H ( j 0 ) = 1
= H1
2
H 1 ( j 0 ) = H 1
2
( )2
dH d
63
d d
dominant
2
dH d
d H d
+ H
d d
Orthogonal
At resonance, the phase changes much faster than magnitude, and |H|=1 near resonance. ignore amplitude noise and AM to PM conversion as well.
dH d
d = d
N out ( s ) N in ( s )
H1
( )
2
d d
0 d Q= 2 d
N out ( s ) N in ( s )
4 Q 2 ( )
H 1 02
2
64
4 Q 2 ( )
02
H1 0 = (2 Q )
N in ( s ) 2P S
Ps is the signal power at active device input. input If the transistor and bias were noiseless, then the only noise present would be due to the resonator losses. The transistors and the bias will add noise to the minimum noise of
N in ( s )
= kT
65
L C Q1 Q2
O ill t m Oscillator
Ibias
= kT +
2 i nt Rp
2 + i cn R p (1 )
F =1+
2 kT
i R p (1 ) kT
H 1 0 FkT PN = (2 Q ) 2 Ps
66
67
3. Crystal/CP Intercept
Div id
er n
oi s e
Cr y stal n
C P nois e
oise
2. CP/VCO Intercept
nois e
no ise
68
Div ider
nois e
Crys ta
l noi
se
To
ta l
no
ise
CP ise no
PD noise
O VC
ise no ise no
se noi
F LP
69
-80 -90 -100 -110 -120 -130 -140 -150 -160 0.1 1.0 10 100 1000
Parameter C1
10000
C2 R
70
Testing
Definition: Having designed and fabricated a device, testing must determine whether or not the device is free from any manufacturing defect. Testing is distinctly different from verification, which checks the correctness of the design. Forms of testing: g Production testing Characterization testing
72
Production Testing
Applied to every manufactured device Major j considerations Reduce cost; minimize test time per device. Maximize quality; reduce defect level (DL), defined as fraction of bad devices passing test. Reference M. L. Bushnell and V. D. Agrawal, Essentials of Electronic Testing for Digital, Memory & Mixed-Signal VLSI Circuits, Boston: Springer, Springer 2000, 2000 Chapter 3 3.
73
User Interface
74
Characterization Testing
Performed at the beginning of production phase. Objective: j To verify y the design, g , manufacturability, y, and test program. Method: Few devices tested very thoroughly g Failures are often diagnosed Tests are more elaborate than the production tests Test time (and testing cost) not a consideration Test program is verified and corrected in necessary ATE system and additional laboratory setup may be used
76
RF Tests
Basic tests Scattering parameters (S-parameters) Frequency and gain measurements Power measurements Power P efficiency ffi i measurements t Distortion measurements Noise measurements
77
78
a1 Port 1 (input) b1 Input return loss Output return loss Gain Isolation RF Device
a2 Port 2 (output) b2 S11 = b1/a1 S22 = b2/a2 S21 = b2/a1 S12 = b1/a2
79
DUT a1 Digitizer b1
Directional couplers
a2 Digiti er Digitizer b2
80
81
Power Measurements
Receiver
Minimum detectable RF power Maximum M i allowed ll di input power Power levels of interfering tones
Transmitter
Maximum RF power output Changes in RF power when automatic gain control is used RF power distribution over a frequency band Power-added efficiency (PAE)
84
Spur Measurement
Spur is a spurious or unintended frequency in the output of an RF device. Example: leakage of reference frequency used in the phase detector of PLL. A spur can violate the channel interference standard of a communication system. Complete power spectrum measured in characterizing phase to determine which interfering frequencies should be checked during production testing.
RF power s R spectrum (dBm/M MHz) 10 40 80 0 200 400 600 800 MHz 1000 1200 1400
85
SPUR
Harmonic Measurements
Multiples of the carrier frequency are called harmonics. Harmonics are g generated due to nonlinearity y in semiconductor devices and clipping (saturation) in amplifiers. Harmonics may interfere with other signals and must be measured to verify that a manufactured device meets the specification. p
86
87
PAE Example
Following measurements are obtained for an RF power amplifier: RF Input power = +2dBm RF output power = +34dBm DC supply voltage = 3V DUT current = 2.25A PAE is calculated as follows: PRF(input) = 0.001 102/10 = 0.0015W PRF(output) = 0 0.001 001 1034/10 = 2.5118W 2 5118W Pdc = 3 2.25 = 6.75W PAE = (2.5118 (2 5118 0.00158)/6.75 0 00158)/6 75 = 0.373 0 373 or 37 37.2% 2%
89
Ideal
Time (s)
400
600
Power (dBm)
Actual measurement
Overshoot Missing gain step Nonlinearity
200
Time (s)
400
600
91
92
RF Communications Standards
Frequency range Channel F Ch l (MHz) bandwidth (MHz) 802.11b (WLAN) 2400-2500 22 16.8 802 11 / (WLAN) 2400 802.11a/g 2400-2500 2500 (g) ( ) 5000-6000 (a) 802 16a (WIMAX) 802.16a 2000-11000 2000 11000 3 most common bands: 2500, 3400, 5800 3100-10600 3 bands: 890-960 1710-1880 1850 1990 1850-1990 450, 800, 1700, 1900, 2100 2,400-2,500 Data D t rate t (Mbps) 11 54 M d l ti format Modulation f t CCK OFDM, 52 OFDM subcarriers (4 pilots, 48 data channels) OFDM, 256 OFDM Subcarriers(200 actually used; 192 are data channels OFDM GMSK CDMA FSK
93
slope = a x f(x) = ax
96
97
98
0 Is
99
101
Harmonic Distortion
Harmonic distortion is the presence of multiples of a fundamental frequency of interest. N times the fundamental frequency is called Nth harmonic. Disadvantages: Waste of power in harmonics. Interference from harmonics. Measurement: Single-frequency input signal applied. Amplitudes of the fundamental and harmonic frequencies are analyzed to quantify distortion as: Total harmonic distortion (THD) Signal, noise and distortion (SINAD) 102
103
Also, THD(dB) = 10 log THD(%) For an ideal distortionless signal, signal THD = 0% or dB
104
THD Measurement
THD is specified typically for devices with RF output. The fundamental and harmonic frequencies q together g are often beyond the bandwidth of the measuring instrument. Separate power measurements are made for the fundamental and each harmonic. THD is tested at specified power level because THD may be small at low power levels. Harmonics appear pp when the output p p power of an RF device is raised.
105
106
107
Gain Compression
The harmonics produced due to nonlinearity in an amplifier reduce the fundamental frequency power output (and gain). This is known as gain compression. As input power increases, so does nonlinearity causing greater gain compression. A standard measure of Gain compression p is 1-dB compression p point power level P1dB, which can be Input referred for receiver, or Output referred for transmitter
108
Amplitude A e
time
Amplitude A e
time
f1
frequency
f1
frequency
109
Amplitude A e
time
Amplitude A e
time
f1
frequency
f1
f2
f3
frequency
110
Effect of Nonlinearity
Assume a transfer function, vo = a0 + a1 vi + a2 vi2 + a3 vi3 Let vi = A cos t Using the identities ( = 2f): cos2 t = (1 + cos 2t)/2 cos3 t = (3 cos t + cos 3t)/4 We get, vo = a0 + a2A2/2 + (a1A + 3a3A3/4) cos t +( (a2A2/2) ) cos 2t + ( (a3A3/4) ) cos 3t
111
112
1 dB
P1dB(output) )
1 dB Compression point
Compression region
113
114
115
Intermodulation Distortion
Intermodulation distortion is relevant to devices that handle multiple frequencies. Consider an input signal with two frequencies 1 and 2: vi = A cos 1t + B cos 2t Nonlinearity in the device function is represented by vo = a0 + a1 vi + a2 vi2 + a3 vi3 Therefore, device output is vo = a0 + a1 (A cos 1t + B cos 2t) + a2 (A cos 1t + B cos 2t)2 + a3 (A cos 1t + B cos 2t)3 DC and fundamental 2nd order terms 3rd order terms
116
Problems to Solve
Derive the following: vo = a0 + a1 ( (A cos 1t + B cos 2t) ) + a2 [ A2 (1+cos 1t)/2 + AB cos (1+2)t + AB cos (1 2)t + B2 (1 (1+cos cos 2t)/2 ] + a3 (A cos 1t + B cos 2t)3 Hint: Hi t Use U the th identity: id tit cos cos = [cos( + ) + cos( )] / 2 Simplify Si lif a3 (A cos 1t + B cos 2t)3
117
2 3 4 5 6 7 N
2 2 2 2 2 2 2
2 4 6 8 10 12 2N 2
4 6 8 10 12 14 2N
2f1 , 2f2 3f1 , 3f2 4f1 , 4f2 5f1 , 5f2 6f1 , 6f2 7f1 , 7f2
f1 + f2 , f2 f1 2f1 f2 , 2f2 f1 2f1 2f2 , 2f2 2f1 , 3f1 f2 , 3f2 f1 3f1 2f2 , 3f2 2f1 , 4f1 f2 , 4f2 f1 3f1 3f2 , 3f2 3f1 , 5f1 f2 , 5f2 f1 , 4f1 2f2 , 4f2 2f1 4f1 3f2 , 4f2 3f1 , 5f1 2f2 , 5f2 2f1 , 6f1 f2 , 6f2 f1
Nf1 , Nf2 . . . . .
118
Problem to Solve
Write Distortion products for two tones 100MHz and 101MHz Harmonics Od Order I t Intermodulation d l ti products d t (MH (MHz) ) (MHz) 2 200, 202 1, 201 3 300, 3003 99, 102, 301, 302 4 400, 404 2, 199, 203, 401, 402, 403 5 500, 505 98, 103, 299, 304, 501, 503, 504 6 600, 606 3, 198, 204, 399, 400, 405, 601, 603, 604, 605 97 104, 298, 97, 298 305, 305 499 499, 506 506, 701 701, 707 707, 703 703, 7 700, 707 704, 705, 706 Intermodulation products close to input tones are shown in bold.
119
Ampli itude
DUT
frequency
f2 f1
f1 f2
Ampli itude
f1 f2 2f1 2f2
frequency
120
Amplitu ude
DUT
f1 f2
2f1 f2 Amplitud de 2f2 f1 frequency
f1 f2
3f1 3f2
121
Problem to Solve
For A = B, i.e., for two input tones of equal magnitudes, show that: Output amplitude of each fundamental frequency, f1 or f2 , is 9 a1 A + a3 A3 4 Output amplitude of each third-order intermodulation frequency, 2f1 f2 or 2f2 f1 , is 3 a3 A3 4
122
a1 A
3a3 A3 / 4 A
123
IP3
IP3 Graph
127
Cross Modulation
Cross modulation is the intermodulation distortion caused by multiple carriers within the same bandwidth. Examples:
In cable TV, same amplifier is used for multiple channels. Orthogonal frequency division multiplexing (OFDM) used in WiMAX or WLAN use multiple carriers within the bandwidth of the same amplifier.
Measurement:
Turn on all tones/carriers except one Measure the power at the frequency that was not turned on
B. Ko, et al., A Nightmare for CDMA RF Receiver: The Cross Modulation, Proc. 1st IEEE Asia Pacific Conf. on ASICs, Aug. 1999, pp. 400-402.
128
What is Noise?
Noise in an RF system is unwanted random fluctuations in a desired signal. Noise is a natural phenomenon and is always present in the environment. Effects of noise: Interferes with detection of signal g ( (hides the signal). g ) Causes errors in information transmission by changing signal. Sometimes noise might imitate a signal falsely. All communications system design and operation must account for noise.
130
Describing Noise
Consider noise as a random voltage or current function, x(t), over interval T/2 < t < T/2. Fourier transform of x(t) is XT(f). Power spectral density (PSD) of noise is power across 1 Sx(f) = lim [ E{ |XT(f)|2 } / (2T) ] T This is also expressed in dBm/Hz. volts2/Hz
131
Thermal Noise
Thermal (Johnson) noise: Caused by random movement of electrons due to thermal energy that is proportional to t temperature. t Called white noise due to uniform PSD over all frequencies. Mean square open circuit noise voltage across R resistor [Nyquist, 1928]: v2 = 4hfBR / [exp(hf/kT) 1] Where Planks constant h = 6.626 1034 J-sec Frequency and bandwidth in hertz = f, B Boltzmanns B lt constant t t k = 1.38 1 38 10 23 J/K Absolute temperature in Kelvin = T
132
Problem to Solve
Given that for microwave frequencies, hf << kT, derive the following Rayleigh-Jeans approximation: v2 = 4kTBR
Show that at room temperature (T = 290K), thermal noise power supplied by resistor R to a matched load is ktB or 174 dBm/Hz.
R R
Matched load
133
Measuring Noise
Expressed as noise power density in the units of dBm/Hz. Noise sources: Resistor at constant temperature, noise power = kTB W/Hz. Avalanche diode Noise temperature: Tn = (Available noise power in watts)/(kB) kelvins Excess noise ratio (ENR) is the difference in the noise output between hot ( (on) ) and cold (off) ( ) states, , normalized to reference thermal noise at room temperature (290K): ENR = [k( Th Tc )B]/(kT0B) = ( Th / T0) 1 Where noise output in cold state is takes same as reference. 135 10 log ENR ~ 15 to 20 dB
Po ower (dBm m)
So/No
Frequency (Hz)
136
137
Fsys
F1
138
Y Factor
Y factor is the ratio of output noise in hot (power on) state to that in cold (power off) state. Y = = Nh / Nc Nh / N0
Y is a simple ratio. Consider, C id Nh = kThBG and d Nc = kT0BG Then Nh Nc = kBG( Th T0 ) or kBG = ( Nh Nc ) / ( Th T0 ) Noise factor, F = Nh /( kT0 BG) = ( Nh / T0 ) [ 1 / (kBG) ] = ( Nh / T0 ) ( Th T0 ) / (Nh Nc ) = ENR / (Y 1)
140
Calibration
Noise source ENR Tester (power meter) F2, Y2
Y2 = Nh2 / Nc2 2, where Nh2 = measured power for hot source Nc2 = measured p power for cold source F2 = ENR / (Y2 1)
142
Y12 = Nh12 / Nc12, where Nh12 = measured power for hot source Nc12 = measured power for cold source F12 = ENR / ( Y12 1 ) G1 = ( Nh12 Nc12 ) / ( Nh2 Nc2 )
143
Problem to Solve
Show that from noise measurements on a cascaded system, the noise factor of DUT is given by F2 1 F1 = F12 G1
144
Phase Noise
Phase noise is due to small random variations in the phase of an RF signal. In time domain, phase noise is referred to as jitter. Understanding phase:
amplitude noise t phase noise t
V sin t
Frequency (rad/s)
Frequency (rad/s)
145
146
[V + (t)] sin [t + (t)] = [V + (t)] [sin t cos (t) + cos t sin (t)] [V + (t)] sin t + [V + (t)] (t) cos t In-phase carrier frequency with amplitude noise White noise (t) corresponds to noise floor Quadrature-phase carrier frequency with amplitude and phase noise Short-term phase noise corresponds to phase noise spectrum
147
DUT
offset Spectrum analyzer power measurement Power (dBm) over resolution bandwith (RBW)
Hz
carrier
149
Problem to Solve
Consider the following spectrum analyzer data: RBW = 10Hz Frequency offset = 2kHz Pcarrier = 3.31 dBm Poffset = 81.17 81 17 dBm dB Determine phase noise in dBc/Hz at 2kHz from the carrier.
151
153
Problem to Solve
Find the testing cost for a good device shipped using the data given in the previous slide.
155
Testing Cost
Tester time per year: T = 365 24 3600 0.6 Number of devices tested per year: NT = T/(1.5 + 1.0) = 18,921,600 s = 7,568,640
Number of good devices produced per year: N = NT Yield = 7,568,640 0.8 = 6,054,912 Testing cost per year: (1,000,000 + 250,000)/5 ) C = ( Testing cost per device shipped: Cost = C/N = 4.13 cents
156
= 250,000 dollars
157
LO
Duplexer r
90 ADC
TA
0
PA Phase Splitter
LO
DAC
90
SOC RF
DAC
BASEBAND
0 0
Tester
Digital pin or digitizer RF detector and buffer amplifier
Computer
Filter
160
References
SOC BIST
J. Dabrowski, BiST Model for IC RF-Transceiver Front-End, Proc. 18th IEEE International I t ti l Symp. S on Defect D f t and dF Fault lt Tolerance T l in i VLSI S Systems t , 2003. D. Lupea, et al., RF-BIST: Loopback Spectral Signature Analysis, Proc. Design, Automation and Test in Europe Conf., 2003.
Low-cost testing
F. Goh, et al., Innovative Technique for Testing Wide Bandwidth F Frequency Response, R Wireless Wi l Broadband B db d F Forum, Cambridge, C b id UK, UK 2004.
161
Lecture 8: RF BIST
Purpose
Develop Built-In Self-Test (BIST) approach using direct digital synthesizer (DDS) for functionality testing of analog circuitry in mixed-signal systems Provides BIST-based measurement of Amplifier A lifi linearity li it (IP3) Gain and frequency response Implemented in hardware IP3, gain, and freq. response measured
163
Outline
Overview of direct digital synthesizers (DDS) 3rd order inter-modulation p product ( (IP3) ) BIST architecture Test pattern generator Output response analyzer Experimental results Implementation in hardware IP3 Measurements
164
165
Digital Circuits Sine to-A R D-toAccum W Lookup Conv. -ulator Table 1/fout 1/fout 1/fout Low Pass Filter
fclkFr fout= N 2
Sine Wave
Fr
clk
1/fout
1/fclk
1/fclk
1/fclk
166
Intermodulation
f1 f2 f2- f1 f1 f2 2f1- f2 2f2- f1 f1+f2 2f1 2f2 3f1 3f2
7 8 freq
0 2 4 6 8 10 12 14 16 18 20 22 24 freq
IM3 Two signals with different frequencies are applied to a nonlinear system Output exhibits components that are not harmonics of input fundamental frequencies Third-order intermodulation (IM3) is critical Very close to fundamental frequencies
Analog and mixed signal testing, FDAI, 2008
167
Mathematical Foundation
Input 2-tone: x(t)=A1cos 1t + A2 cos 2t Output of non-linear device: y(t)=0+1x(t)+2x2(t)+3x3(t)+ Substituting x(t) into y(t):
y(t) = 2(A12+A22) + [1A1+3A1(A12+2A22)]cos1t + [1A2+3A2(2A12+A22)]cos2t + 2(A12cos21t+A22cos22t ) + 2A1A2[cos(1+2)t+cos(1-2)t] + 3[A13cos31t+A22cos32t] + 3{A12A2[cos(21+2)t+cos(21-2)t] +A1A22[ [cos(2 ( 2+1)t+cos(2 ( 2-1)t]}
Analog and mixed signal testing, FDAI, 2008
1A P 3A2
freq
168
21-2 1 2 22-1
IM3 IP3
P/2
169
2-tone Waveform
170
171
171
y(t)
accumulator
172
DC1 Accumulator
y(t) x f2 DC1 A221 y(t) Ripple in slope due to low frequency components Longer accumulation f2 reduces effect of ripple
X DC1
1A 3A2
freq
Simulation Results
Millions 140 120 100 80
173
DC2 Accumulator
y(t) x f2 DC2 3/8A12A223 Ripple is bigger for DC2 Test controller needed to 2f2-f1 obtain DC2 at integral multiple p of 2f2-f1
y(t) DC2 X
1A 3A2
freq
Simulation Results
f2 2f2-f1
Millions 20
15
10
174
BIST-based P Measruement
DC1 & DC2 same proportionality to power at f2 & 2f2-f1 Only need DC1 & DC2 from accumulators to calculate P = 20 log (DC1) 20 log (DC2)
Simulation Results
70 60 50 40 30 20 10 0 1
21
41
61
81 101 121 141 161 181 201 221 241 261 Clock cycles (x50)
175
BIST Architecture
BISTBIST -based IP3 measurement Reduce circuit by repeating test sequence for DC2 BISTBIST -based Gain & Frequency Response is subset
x(t )=cos cos( (f2) x(t)= )=cos cos( (f1)+ )+cos cos( ()= f2), f1 f2 2f2-f1
DAC
DUT
ADC
Accum
LUT1
y(t)
Output Response Analyzer
X X
Accum
Accum
LUT3
Accum
FPAA DUT
177
Hardware Results
30
BIST measures P 14
Spectrum analyzer
25 5 20 15 10 5
Percentage(%)
P14
201
401
601
Measured Delta_P(dB)
178
Spectrum analyzer
D elt ta P (dB) )
P22
501
1001
1501
2001
2501
2 1.5 1 0.5 0
16.30 18.94 21.58 24.23 26.87 29.51
179
Conclusion
BIST-based approach for analog circuit functional testing DDS-based TPG Multiplier/accumulator-based M lti li / l t b d ORA Good for manufacturing or in-system circuit characterization and on-chip compensation Amplifier linearity (IP3) Gain and frequency response Measurements with hardware implementation Accurately A t l measures IP3 30dB Measurements of IP3 > 30dB imply higher values
181