Digitally-Assisted Compensation for Timing Skew in ATE Systems
K. Asami T. Tateiwa T. Kurosawa H. Miyajima H. Kobayashi
Advantest Corporation Gunma University
IEEE International Mixed-Signals, Sensors, and Systems Test Workshop
May 2011
Contents
• Research Goal
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Considerations – Window
– Gain Adjustment
• Application
Timing skew is a major problem in ATE systems
Digital compensation for timing skew
⇒ Linear phase is important
Conventional linear-phase digital filter ⇒ coarse timing adjustment Proposed linear-phase digital filter ⇒ fine timing adjustment
Research Goal
Fine time resolution
Linear phase
Features of Proposed Digital Filter
τ Fine time shift
F(t)
F(t-τ)
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
• Conclusion
Linear Phase FIR Filter Impulse Response
0 3 6
0 3
6
0 3 7
0 3
7 (1)Case 1
odd # of taps ・even symmetry
(2) Case 2
even # of taps・even symmetry
4 4
Frequency Characteristics
Case 1
Case 2
Case 3
Case 4
Phase : proportional to ω (linear phase) Time resolution of group delay Ts/2
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
Ideal LPF
:Sampling Frequency Frequency Characteristics
1.0
Impulse Response
Fourier Transform
1 2 3 4 5 -5 -4 -3 -2 -1
Discrete-Time Representation of Ideal LPF
Fourier Transform
FIR (Finite Impulse Response)
zero
Impulse Response Time-Shift
1 2 3 4 5 -5 -4 -3 -2 -1
No change of Gain
Δt time-shift of impulse response
Time-Shift and Filter Coefficients
FIR filter
1 2 3 4 5 -5 -4 -3 -2 -1
IIR Filter
Time Shift
Ideal Delay-Filter
zero
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
• Conclusion
2-Tap Filter: Model
1 0
2-Tap Filter: Delay Model
0 1
1 2 3 4 5 -5 -4 -3 -2 -1
FIR
IIR
2-Tap Filter: Delay Model
0 1
IIR FIR
Proposed Delay Digital Filter
(a) FIR Filter (b) Ideal Delay Filter
(c) Delay Digital Filter
Window
Frequency Characteristics of Proposed Delay Digital Filter
Case 1
Case 2
Case 3
Case 4
( )
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
• Conclusion
Comparison of 2-Tap Filter Impulse Responses
5 10 15 20 25 30
-0.2 0 0.2 0.4 0.6
2 taps FIR Filter
5 10 15 20 25 30
-0.2 0 0.2 0.4 0.6
Delayed Filter (0.3 samples delay)
2-tap FIR coefficients
Impulse response changes.
Proposed Delay Filter
(
0.3 samples delay)
2-Tap FIR Filter
zero
Non-zero
Comparison of
2-Tap Filter Frequency Characteristics
-0.5 0 0.5
-20 -15 -10 -5 0
Gain [dB]
-0.5 0 0.5
-2 0 2
Normalized Frequency (Fs=1.0)
Phase [radian]
No change of gain
Phase slope changes
Proposed delay filter
Original filter Original filter and proposed delay filter
Hann
Finite Tap Truncation of Proposed Delay Filter
10 20 30 40 50 60
-0.2 0 0.2 0.4 0.6
61 taps Cosine Roll-off Filter
0.2 0.4 0.6
Delayed Filter with Hann window (0.3 samples delay)
61
タップ コサインロールオフフィルタ
10 20 30 40 50 60
-0.2 0 0.2 0.4 0.6
61 taps Cosine Roll-off Filter
10 20 30 40 50 60
-0.2 0 0.2 0.4 0.6
Delayed Filter (0.3 samples delay)
Delay Filter
(
0.3 samples delay)
61-Tap Cosine Roll-off FilterDelay Filter
(
0.3 samples delay)
Rectangular window
Effects of Window
0 0.05 0.1 0.15 0.2 0.25
-10 -5 0
Gain [dB]
0 0.05 0.1 0.15 0.2 0.25
30.299 30.2995 30.3 30.3005 30.301
Normalized Frequency (Fs=1.0)
Group Delay [samples]
Hann window
Rectangular window
Frequency characteristics of delay filter with 61-tap truncation
Gibbs oscillation of group delay
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
How to Apply Window
1 2 3 4 5 -5 -4 -3 -2 -1
Δt Ts
1 2 3 4 5 -5 -4 -3 -2 -1
Δt Ts
t Ts
t Ts
Window
Centered at origin
Centered at
impulse response center
Window center shifted by Δt
0 0.1 0.2 0.3 0.4 0.5 -100
-50 0
Power [dB]
マルチレート・フィルタの周波数特性(非対称)
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Normalized frequency
Phase [rad]
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Power [dB]
マルチレート・フィルタの周波数特性(対称)
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Normalized frequency
Phase [rad]
Frequency Characteristics of Delay Filter after Applying Window
Delay 0.3 samples Filter Tap 100 taps
Window centered at origin Window centered at impulse response
Gain[dB]Phase[degree]
0.1 0.15 0.2 0.25 0.3 0.35 0.4 49.79
49.795 49.8 49.805
Normalized frequency
Group delay
群遅延特性(非対称)
0.1 0.15 0.2 0.25 0.3 0.35 0.4 49.79
49.795 49.8 49.805
Normalized frequency
Group delay
群遅延特性(対称)
Group Delay Characteristics of Delay Filter after Applying Window
Window centered at origin Window centered at impulse response
Delay 0.3 samples Filter Tap 100 taps
Window Han
Pass band (0.1~0.4)・Fs
0 0.1 0.2 0.3 0.4 0.5 -100
-50 0
Power [dB]
マルチレート・フィルタの周波数特性(非対称)
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Normalized frequency
Phase [rad]
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Power [dB]
マルチレート・フィルタの周波数特性(対称)
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Normalized frequency
Phase [rad]
Frequency Characteristics of Delay Filter after Applying Window
gain[dB]Phase[degree]
Delay 0.3 samples Filter Tap 100 taps
Window centered at origin Window centered at impulse response
0.05 0.1 0.15 0.2 0.25 0.3 49.79
49.795 49.8 49.805
Normalized frequency
Group delay
群遅延特性(非対称)
0.05 0.1 0.15 0.2 0.25 0.3
49.79 49.795 49.8 49.805
Normalized frequency
Group delay
群遅延特性(対称)
Group Delay Characteristics of Delay Filter after Applying Window
Delay 0.3 samples Filter Tap 100 taps
Window Han
Pass band (0.05~0.3)・Fs
Window centered at impulse response Window centered at origin
0.05 0.1 0.15 0.2 0.25 0.3 49.79
49.795 49.8 49.805
Normalized frequency
Group delay
群遅延特性(非対称)
0.05 0.1 0.15 0.2 0.25 0.3
49.79 49.795 49.8 49.805
Normalized frequency
Group delay
群遅延特性(対称)
Group Delay Characteristics of Delay Filter after Applying Window
Delay 0.3 samples Filter Tap 100 taps
Window centered at impulse response Window centered at origin
Applying window centered at impulse response
Constant group delay over entire passband
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
• Conclusion
Proposed Filter DC Gain Adjustment
Digital filter DC gain :
1 2 3 4 5 -5 -4 -3 -2 -1
t Ts
Δt Ts
ana0 a1
a2 a3
a4 a5
a6
a7 a8
a9
a10
DC gain adjustment due to finite tap truncation is required
● Original filter
● Delay filter
N0 0.1 0.2 0.3 0.4 0.5 -100
-50 0
Normalized Frequency
Power [dB]
ゲイン未調整
0 0.1 0.2 0.3 0.4 0.5
-150 -100 -50 0
Frequency
Phase [degree]
0 0.1 0.2 0.3 0.4 0.5
-100 -50 0
Normalized Frequency
Power [dB]
ゲイン未調整
0 0.1 0.2 0.3 0.4 0.5
-150 -100 -50 0
Frequency
Phase [degree]
Frequency Characteristics of Proposed Delay Filter
With DC gain adjustment WithoutDC gain adjustment
Delay 0.3 samples
Filter Tap 101 taps
Window Han
Cut-off Freq. 0.4・Fs
0 0.1 0.2 0.3 0.4 0
0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Normalized Frequency
gain [dB]
マルチレート・フィルタの周波数特性
0 0.1 0.2 0.3 0.4
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Normalized Frequency
gain [dB]
マルチレート・フィルタの周波数特性
Gain Characteristics of Proposed Delay Filter
Delay 0.1samples
Filter Tap 101 taps
Delay 0.3 samples Filter Tap 101 taps
0.005dB 0.0006dB
With DC gain adjustment WithoutDC gain adjustment
gain[dB]
0 0.1 0.2 0.3 0.4 0
0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Normalized Frequency
gain [dB]
マルチレート・フィルタの周波数特性
Original FIR filter
With DC gain adjustment WithoutDC gain adjustment
Gain Characteristics of Proposed Delay Filter
0 0.1 0.2 0.3 0.4
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Normalized Frequency
gain [dB]
マルチレート・フィルタの周波数特性
Delay 0.1samples
Filter Tap 101 taps
Window Han
Cutoff Freq. 0.4・Fs
Delay 0.3 samples Filter Tap 101 taps
Window Han
Cutoff Freq. 0.4・Fs
0 0.1 0.2 0.3 0.4 0
0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Normalized Frequency
gain [dB]
マルチレート・フィルタの周波数特性
Original FIR filter
With DC gain adjustment WithoutDC gain adjustment
Gain Characteristics of Proposed Delay Filter
0 0.1 0.2 0.3 0.4
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Normalized Frequency
gain [dB]
マルチレート・フィルタの周波数特性
Delay 0.1samples
Filter Tap 101 taps
Delay 0.3 samples Filter Tap 101 taps
DC gain adjustment
Delay filter gain
Original FIR filter gain
Contents
• Research Purpose
• Conventional Linear Phase Digital Filter Condition
• New Linear Phase Digital Filter Condition
– Time-Shift, Impulse Response of Ideal Filter – New Linear Phase Digital Filter
• MATLAB Simulation
• Design Consideration – Window
– Gain Adjustment
• Application
• Conclusion
I/Q Delay Mismatch in Quadrature Modulator
s(t) t
p/2 fc
DAC
DAC I(t) = cos (2pf0t)
Q(t) = sin(2pf0t) SSB signal input
fc I(t)+jQ(t)
0 fc- f0 fc+f0
f f
delay
in analog domain
Image rejection ratio
I/Q Delay Mismatch Compensation in Quadrature Modulator
s(t) t
p/2 fc
DAC
DAC I(t) = cos (2pf0t)
Q(t) = sin(2pf0t) SSB signal
DAC : digital-to-analog converter SSB : single side band
fc I(t)+jQ(t)
0 fc- f0 fc+f0
f f
Delay
in analog domain
digital timing compensation τ
Matlab Simulation Results
(b) Timing skew case
Delay 0.3 samples Filter tap # 61 taps
-0.5 0 0.5
-140 -120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
タイミング・スキューのある信号image signal
-0.5 0 0.5
-140 -120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
理想信号
(a) Ideal case
signal
Matlab Simulation Results
-0.5 0 0.5
-140 -120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
ゲイン無調整
(c) Compensation using delay filter
Without adjustment of window, gain
signal image
-0.5 0 0.5
-140 -120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
ゲイン改善signal image
Delay 0.3 samples Filter tap 61 taps
Window Han
FFT points 1024 points
(d) Compensation using delay filter With adjustment of window, gain
M channel ADCs M-times sampling rate
Interleaved ADC System
ADC :analog-to-digital converter
ADC1 M
U X
CLK1
CLK2
ADC2
Ts
t CLK 1
CLK 2
Timing Skew in Interleaved ADC System
fin Fs- finFs f
fin f 0
0
Analog Input
Fs=1/Ts Fs
fin
Digital output Dout
ADC1 M U
X
CLK1
CLK2
ADC2
Ts
t CLK 1
CLK 2
Timing Skew Compensation in Interleaved ADC System
f f Analog input
fin Clock skew effect
compensation
Digital output Dout
Matlab Simulation Results
0 0.1 0.2 0.3 0.4 0.5
-120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
スキューのある信号
0 0.1 0.2 0.3 0.4 0.5
-120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
スキューのある信号
Signal Signal Spurious
Delay 0.3 samples Filter tap 61 taps
Window Han
FFTpoints 1024 points
(a) Ideal case (b) Timing skew case
0 0.1 0.2 0.3 0.4 0.5 -120
-100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
補正後のスペクトラム
0 0.1 0.2 0.3 0.4 0.5
-120 -100 -80 -60 -40 -20 0
Normalized frequency
Magnitude [dB]
補正後のスペクトラム
Matlab Simulation Results
Delay 0.3 samples Filter tap 61 taps
Signal Spurious Signal Spurious
(c) Compensation using delay filter
Without adjustment of window, gain
(d) Compensation using delay filter With adjustment of window, gain
Conclusion
● Linear phase digital filter
with fine time resolution of group delay
● Design consideration - How to apply window - DC gain adjustment
● Application Examples
- I/Q delay mismatch compensation in quadrature modulator
- Timing skew compensation in interleaved ADC system Future work
● Implementation issues
– Finite word length, finite tap effects – LSI implementation