Академический Документы
Профессиональный Документы
Культура Документы
University of Toronto
slide 1 of 20
D.A. Johns, K. Martin, 1997
V out
+ b2 2
+ + bN 2
(1)
b i equals a 1 or a 0 (i.e. a binary digit). V ref an analog reference; V out output . V out = V ref ( b 1 2
1 2 N
+ b2 2
+ + bN 2
)
N
(2)
University of Toronto
slide 2 of 20
D.A. Johns, K. Martin, 1997
A multiplying D/A allows V ref to be a varying input V out proportional to multiplication of V ref and B in . For ideal D/A , output signal is a well defined value no quantization error!
V out -------V ref
1 3/4 1/2 1/4 0 00 01 10 11 (100)
University of Toronto
slide 3 of 20
D.A. Johns, K. Martin, 1997
Guaranteed monotonic
b2
V out
Delay through the switch network major speed limitation Resistors might be realized using polysilicon If n-channel only used, can be laid out small
b3 b3 b3 b3 b2 b1 b2
Requires 2N resistors
3-bit
University of Toronto
slide 4 of 20
D.A. Johns, K. Martin, 1997
b1 b2 b3
Large cap load on buffer input Can pipeline digital decoding for faster speed Requires 2N resistors
Resistors
V out 3-bit
University of Toronto
slide 5 of 20
D.A. Johns, K. Martin, 1997
Folded-resistor-string D/A
(Abrial, JSSC, Dec. 1988) V ref
word lines
b1 b2
2 to 1 of 4 decoder
Less capacitance load over the single bus approach Requires 2N resistors
bit lines
V out
2 to 1 of 4 decoder
b3 b4
University of Toronto
4-bit
slide 6 of 20
D.A. Johns, K. Martin, 1997
V out
V ref V out
Only N resistors Resistor and current ratios are on the order of 2N No guarantee of monotonicity. Prone to glitches (more later).
University of Toronto
slide 7 of 20
D.A. Johns, K. Martin, 1997
V out 4R
2R
4R 3R
2R
4R
V ref
1 V A = -- ( V ref ) 4
4-bit
Reduced resistor spread Keep repeating this procedure > R-2R ladder
University of Toronto
slide 8 of 20
D.A. Johns, K. Martin, 1997
R 2 R' 2 R R
R 3 R' 3 R
R 4 R' 4 2R
2R
V ref ----------4R
2R
V ref ----------8R
2R
V ref ----------16R
R' 4 = 2 R R 4 = 2 R || 2 R = R R' 3 = R + R 4 = 2 R R 3 = 2 R || R' 3 = R Small size, good matching (only R and 2R)
(4)
University of Toronto
slide 9 of 20
D.A. Johns, K. Martin, 1997
b1
Ir 2R R Ir Ir -2
b2
2R R Ir -4
b3
2R R Ir 4 Ir -8
b4
2R 2R
V out
V ref
Ir 2
Ir 8
4-bit
Currents through the switches are scaled Should scale switch sizes for good accuracy No node voltage changes except for output > fast
University of Toronto
slide 10 of 20
D.A. Johns, K. Martin, 1997
Rf
Vo
b4
I
b3
I
b2
I
b1
I
-5V
4-bit
Node voltages change slower circuit No need to scale switch sizes (smaller size)
University of Toronto
slide 11 of 20
D.A. Johns, K. Martin, 1997
Glitches
Different delays for switching the different currents MSB change often worst case I1 t I2 t I1 + I2 t Glitches can be minimized by limiting the bandwidth but that slows down circuit Use thermometer code to reduce glitches
University of Toronto
slide 12 of 20
D.A. Johns, K. Martin, 1997
Rf V out 1 I1 2 I2
Charge-Redistribution SC D/As
Programmable SC gain amplifier. 16 C 8C 4C 2C C b1 b2 b3 b4
2 ( 1 ) 1 1 2
V out
V ref
1 ( 2 )
C2
4-bit
Sign bit realized by interchanging input phases Carefully clock-waveforms required to minimize voltage dependency of clock-feed-through. Digital codes should be changed when input side of capacitors are connected to ground. Requires extra digital complexity.
University of Toronto
slide 13 of 20
D.A. Johns, K. Martin, 1997
b1
0 0 0 0 1 1 1 1
Binary
b2
0 0 1 1 0 0 1 1
b3
0 1 0 1 0 1 0 1
b2
d1
0 0 0 0 0 0 0 1
b3
d2
0 0 0 0 0 0 1 1
Thermometer Code
d3
0 0 0 0 0 1 1 1
d4
0 0 0 0 1 1 1 1
d5
0 0 0 1 1 1 1 1
d6
0 0 1 1 1 1 1 1
d7
0 1 1 1 1 1 1 1
b1
V out
ref
University of Toronto
slide 14 of 20
D.A. Johns, K. Martin, 1997
C2
2 N unit sized caps Guaranteed monotonic Much lower glitching Low DNL
University of Toronto
slide 15 of 20
D.A. Johns, K. Martin, 1997
Current-Mode D/As
Vout
Thermometer-code High-speed, output feeds directly to resistor Important that delay to all the switches are equal. Overlapped clocks much better than having nonoverlapped clocks.
Col.
Vout
Column Decoder
di
Row
di
Row Decoder
di
I - Src Array
Bias
University of Toronto
slide 16 of 20
D.A. Johns, K. Martin, 1997
V V
d bias
1 Q 1 Q 2 V
2 V
ref R ref
bias
bias
V out
50
University of Toronto
slide 17 of 20
D.A. Johns, K. Martin, 1997
Segmented D/A
Schoeff, 79; Saul, 85; Grebene, 84
R/2 + Vout
and
Very popular
2 MSBs
4 Bit
University of Toronto
slide 18 of 20
D.A. Johns, K. Martin, 1997
Each current source is with a I ref calibrated single reference 64 used so D/A can continue operating
Id1
Id2
Id3
Id4
Id5
I 64
Dynamic technique with current switching for realizing very well-matched current sources Up to 16 bit accuracy
University of Toronto
slide 19 of 20
D.A. Johns, K. Martin, 1997
I out Id1
S1
I ref
S2
I out Id1
0.9Iref Q1
Current source 0.9I added so a low gm device used (W/L equal to 10/75) Re-calibrate before leakage causes 0.5LSB error
Cgs S1 Q1 0.9Iref
Cgs
Calibration
Regular Usage
Minimize clock-feedthrough and charge-injection by having capacitance Cgs and bias voltage V GS large Implies voltage error causes less current deviation.
University of Toronto
slide 20 of 20
D.A. Johns, K. Martin, 1997