Unformatted text preview: (z ) = b1 + b2(z − 1) + b3(z − 1) + · · ·
D (z ) • The numerator of ST F (z ) is arbitrary, but has an order that is one less than D (z ).
• The ST F (z ) does not contain signiﬁcant peaking.
• Each integrator output contain signiﬁcant amounts of the input signal as well as
ﬁltered quantization noise.
If c1 = 0 and c2 = 0, the poles of L(z ) can be moved away from z = 1 along the unit
circle.
Oversampling 2625 Analog ICs; JiehTsorng Wu Stability of SingleStage HighOrder Modulators
Quantizer Im(z) e(k)
x(k) G(z) α y(k)
Re(z) F(z) αG (z )
1
Y (z ) =
· X (z ) +
· E (z )
1 + αG (z )F (z )
1 + αG (z )F (z )
• A modulator is called stable, if the input to the quantizer does not become overloaded,
i.e., e(k ) ≤ ±∆/2.
• All highorder modulators (N > 2) are conditionally stable.
• Modulators with multibit quantizer and DAC exhibit improved stability.
Oversampling 2626 Analog ICs; JiehTsorng Wu Stability of SingleStage HighOrder Modulators
For singlestage modulators with onebit quantizer and DAC:
• As a general rule of thumb, stability can be achieved by keeping NT F e jω ≤ 1.5. • A modulator can be made more stable by placing the poles closer to the zeros in
NT F (z ). But, the SNR is also degraded since the outofband gain of NT F (z ) is also
reduced.
• Stability is also related to the input signal level. Typically want 50–80% of ∆ for stable
input range.
• “Signal overload” and “power on” may cause a conditionally stable modulator to
oscillator. Need additional mechanism to detect instability and force the loop
becoming stable. Oversampling 2627 Analog ICs; JiehTsorng Wu MultiStage Cascaded Modulators
x(k) H 1 y (k)
1 (z) D/A
Error
Cancel y(k) e (k)
1
H 2 y (k)
2 (z) D/A Y1(z ) = S1(z ) · X (z ) + N1(z ) · E1(z )
Oversampling Y2(z ) = S2(z ) · E1(z ) + N2(z ) · E2(z ) 2628 Analog ICs; JiehTsorng Wu MultiStage Cascaded Modulators
The error cancellation logic is
Y (z ) = S2(z ) · Y1(z ) − N1(z ) · Y2(z )
If S2(z ) = S2(z ) and N1(z ) = N1(z ), then Y (z ) = S1(z )S2 (z ) · X (z ) − N1(z )N2 (z ) · E2(z ).
• Also called multistage noise shaping (MASH) architecture.
• Individual loop can be loworder and stable. The resulting noise shaping function
N1(z )N2 (z ) · · · is highorder.
• Sensitive to mismatches between the analog and digital circuitry.
• For loworder loop, the ﬁnite opamp gain can cause noise leakthrough.
• y (k ) has more than one bit, thus complicates the output DAC design in D/A
applications or the decimation ﬁlter design in A/D applications.
Oversampling 2629 Analog ICs; JiehTsorng Wu A ThirdOrder (111) Cascaded Modulators
x(k) z y (k)
1 1 z 2 z 1 y(k) D/A
q 1 (k)
z y (k)
2 1 z y (k)
4 2 z 1 D/A
q 2 (k)
z y (k)
3 1 z 1 D/A Oversampling 2630 Analog ICs; JiehTsorng Wu A ThirdOrder (111) Cascaded Modulators
The outputs of the quantizers are
Y 1 = z − 1 X + 1 − z − 1 E1 Q1 = Y1 − E1 = z −1(X − E1) Y 2 = z − 1 Q 1 + 1 − z − 1 E2 Q2 = Y2 − E2 = z −1(Q1 − E2) Y 3 = z − 1 Q 2 + 1 − z − 1 E3 = z − 2 Q 1 − z − 2 E2 + 1 − z − 1 E3
We have
Y4 = z − 2 Y2 + 1 − z − 1 Y3 = z − 2 Q 1 + 1 − z − 1 2 E3 = z − 3 X − z − 3 E1 + 1 − z − 1 2 E3 and
Y = z − 3 Y1 + 1 − z − 1 Y4 = z − 3 X + 1 − z − 1 Oversampling 2631 3 E3 Analog ICs; JiehTsorng Wu Idle Channel Tones (Pattern Noises)
For a 1storder 1bit modulator and ±∆ = ±1,
y (k ) = sgn[u(k )] = u(k ) + e(k ) u(k + 1) = u(k ) + x (k ) − y (k ) If x (k ) = 0 and u(0) = 0, then
y (k ) = (+1, −1) · · · e(k ) = (+1, 0) · · · If x (k ) = 1/3 and u(0) = 0, then
y (k ) = (+1, −1, +1) · · · e(k ) = (+1, −1/3, +1/3) · · · If x (k ) = 1/2 and u(0) = 0, then
y (k ) = (+1, −1, +1, +1) · · · Oversampling e(k ) = (+1, −1/2, 0, +1/2) · · · 2632 Analog ICs; JiehTsorng Wu Idle Channel Tones (Pattern Noises)
• In above examples, e is periodic and nowhere near white. Diﬀerent initial states just
shift the sequence and the values of e.
• For bounded input u < 1, x is rational ⇔ y is periodic.
• Lowfrequency tones cannot be ﬁltered out by the following decimation ﬁlter.
• Tones also exists in higherorder modulators. The tones might not lie at a single
frequency but instead be shortterm periodic patterns.
• Nearly all types of modulators can produce very highpowered tones near fs /2. Clock
noise near this frequency can couple and demodulate these tones down into the
baseband.
• For ac input, strong peaks and dips in the output noise power may be seen for certain
input frequencies and amplitudes.
Oversampling 2633 Analog ICs; JiehTsorng Wu NoiseShaped Dithering for SingleStage Modulators
d(k)
x(k) y(k) G(z) F(z) • d (k ) is a pseudorandom noise. It is usually generated by a PN sequence generator.
• The power d (k ) must be comparable to that of e(k ). The pdf of d (k ) usually spans
more than ∆/2.
• d (k ) may require 3–8 quantization levels for eﬀective dithering. Oversampling 2634 Analog ICs; JiehTsor...
View
Full
Document
This note was uploaded on 03/26/2013 for the course EE 260 taught by Professor Choma during the Winter '09 term at USC.
 Winter '09
 Choma
 Integrated Circuit, Transistor, The Land

Click to edit the document details