Switched-Capacitor Filters: - "Analog" Sampled-Data Filters
Switched-Capacitor Filters: - "Analog" Sampled-Data Filters
Applications:
Oversampled A/D and D/A converters Analog front-ends (CDS, etc) Standalone filters
E.g. National Semiconductor LMF100
2002 B. Boser 1
Switched-Capacitor Resistor
1 2 S2
Capacitor C is the switched capacitor Non-overlapping clocks 1 and 2 control switches S1 and S2, respectively vIN is sampled at the falling edge of 1
Sampling frequency fS
vIN
S1
vOUT
1 2
T=1/fS
A/D DSP
2002 B. Boser 2
Switched-Capacitor Resistors
The charge transferred from vIN to vOUT each sample period is: Q = C(vIN vOUT)
1 2 S2
vIN
S1
vOUT
The average current flowing from vIN to vOUT is: i = fSC(vIN vOUT)
1 2
T=1/fS
A/D DSP
2002 B. Boser 3
Switched-Capacitor Resistors
i = fSC(vIN vOUT)
With the current through the switched capacitor resistor proportional to the voltage across it, the equivalent switched capacitor resistance is:
2 S2
vIN
S1
vOUT
REQ=
1 fSC 1 2
T=1/fS
A/D DSP
2002 B. Boser 4
Switched-Capacitor Filter
REQ
Lets build an SC filter Well start with a simple RC LPF Replace the physical resistor by an equivalent SC resistor 3-dB bandwidth:
vIN
vOUT C2
1 2 S2
vIN = fS C1 C2
S1
vOUT C2
0 =
C1
REQC2
A/D DSP
2002 B. Boser 5
Switched-Capacitor Filters
In SCFs, all critical frequencies track the sampling frequency
Crystal oscillators for fS are stable to 10ppm/C RC products used in active-RC filters can be tuned, but RCs in active-RC filters dont track together nearly as well
Capacitor ratios in monolithic filters are perfectly stable over time and temperature
Capacitor ratios cant be trimmed easily The trick is to achieve initial ratio accuracies of 1000ppm out of double-poly CMOS processes
A/D DSP
2002 B. Boser 6
Transient Analysis
SC response: extra delay and steps with finite rise time.
Impractical
exaggerated
No problem
A/D DSP
2002 B. Boser 7
Transient Analysis
ZOH
ZOH: pick signal after settling (usually at end of clock phase) Adds delay and sin(x)/x distortion When in doubt, use a ZOH in periodic ac simulations
A/D DSP
2002 B. Boser 8
Periodic AC Analysis
A/D DSP
2002 B. Boser 9
Magnitude Response
1. RC filter output 2. SC output after ZOH 3. Input after ZOH 4. Corrected output (2) over (3) periodic with fs Identical to RC for f<<fs/2
A/D DSP
2002 B. Boser 10
Periodic AC Analysis
SPICE frequency analysis
ac pac linear, time-invariant circuits linear, time-variant circuits
SpectreRF statements
V1 ( Vi 0 ) vsource type=dc dc=0 mag=1 pacmag=1 PSS1 pss period=1u errpreset=conservative PAC1 pac start=1 stop=1M lin=1001
Output
Divide results by sinc(f/fs) to correct for ZOH distortion
A/D DSP
2002 B. Boser 11
A/D DSP
2002 B. Boser 12
A/D DSP
2002 B. Boser 13
Switched-Capacitor Noise
The resistance of switch S1 produces a noise voltage on C with variance kT/C The corresponding noise charge is Q2=C2V2=kTC This charge is sampled when S1 opens
A/D DSP
2 S2
vIN
S1
vOUT
1 2
T=1/fS
2002 B. Boser 14
Switched-Capacitor Noise
The resistance of switch S2 contributes to an uncorrelated noise charge on C at the end of 2 The mean-squared noise charge transferred from vIN to vOUT each sample period is Q2=2kTC
A/D DSP
2 S2
vIN
S1
vOUT
1 2
T=1/fS
2002 B. Boser 15
Switched-Capacitor Noise
The mean-squared noise current due to S1 and S2s kT/C noise is :
i 2 = (Qf s ) = 2k BTCf s2
2
This noise is approximately white (see next slide) and distributed between 0 and fs/2 (noise spectra are single sided by convention) The spectral density of the noise is:
using
REQ
1 = f sC
The noise from an SC resistor equals the noise from a physical resistor with the same value!
A/D DSP
2002 B. Boser 16
fs Sy(f) C
Rsw
3 2 1 0 -1 -2 -3
T/ = 1 T/ = 3
4kBTRsw
-4
0.1
0.4
0.5
and k BTr C
T=
1 fs
A/D DSP
S y ( f )df =
Noise essentially white for T/t > 3 Settling constraints ensure that this condition is usually met in practice Note: This is the noise density of an SC resistor only. The noise density from an SC filter is usually not white.
2002 B. Boser 17
PSS pss period=100n maxacfreq=1.5G errpreset=conservative PNOISE ( Vrc_hold 0 ) pnoise start=0 stop=20M lin=500 maxsideband=10
A/D DSP
2002 B. Boser 18
A/D DSP
Total Noise
Sampled noise in 0 fs/2: 62.2V rms (expect 64V for 1pF)
A/D DSP
2002 B. Boser 20
Low impedance output Can drive R-loads Good for RC filters, OK for SC filters Extra buffer adds complexity, power dissipation
A/D DSP
2002 B. Boser 21
2 voT =
kT C
Rnoise 1 + R switch
2 voT =
kT C
A/D DSP
10 fs 2 fs 2 f s = 8 ... 100 f corner fu f u = 16 ... 200 f corner CT Filter : f u = 50 ... 1000 f corner
A/D DSP
2002 B. Boser 23
SC Filter Summary
Pole and zero frequencies proportional to
Sampling frequency fs Capacitor ratios High accuracy and stability in response Low time constants realizable without large R, C
Amplifier bandwidth requirements comparable to CT filters o Catch: Sampled data filter aliasing
A/D DSP
2002 B. Boser 24