### Essentials Guide to Data Conversion

```The Essential Guide to Data conversion
“Real-World” Sampled Data Systems Consist of ADCs and DACs
ANALOG
DIGITAL
SENSOR
CHANNEL
DSP
g
(
) and Digital-to-Analog
g
Analog-to-Digital
g
g Converters ((DAC))
allow DSPs to interact with real-world signals.
Real-world signals are continuous (analog) signals.
MEMORY
Pressure sensor
Temperature sensor,
sensor etc.
etc
Real-world signal processing allows for efficient and cost effective
extraction of information from a signal.
DIGIT
TAL
VALU
UE
ANALO
OG
AMPLITU
UDE
DAC
TIME
Signal amplitude
Phase, etc.
TIME
QUANTIZED WAVEFORM
DAC RECONSTRUCTED
WAVEFORM
Digital information differs from real-world information in two important
respects…it is sampled, and it is quantized. Both of these restrict how
digital
much
h iinformation
f
ti
a di
it l signal
i
l can contain.
t i
Converter Resolution, INL, and DNL
Converter resolution represents the analog signal at a number of discrete levels
or steps.
The smallest resolvable signal is 1 Least Significant Bit (LSB), which is equal
to FS/8 in this example.
IDEAL
FS
7/8
Nonlinearity
the maximum
deviation
LSBs, ffrom
IIntegral
t
lN
li
it (INL) iis a measure off th
i
d
i ti
iin LSB
a straight line passing through negative full-scale and positive full-scale.
Good INL is required for open-loop systems and many closed-loop systems.
5/8
4/8
1 LSB
3/8
DNL
ANALOG
G
ANALOG
IDEAL
INL
6/8
ACTUAL
QUANTIZATION
UNCERTAINTY
2/8
NON-MONOTONIC
1/8
000 001 010 011 100 101 110
000 001 010 011 100 101 110 111
111
A DAC is monotonic if its output increases or remains the same for an increment
in the digital code, i.e., DNL > –1 LSB (a key requirement in a control system).
Conversely, a DAC is nonmonotonic if the output decreases for an increment
in the digital code.
code
An ADC has no missing codes if the input voltage is swept over the entire input
range and all output code combinations appear at the converter output. A DNL
error of > –0.99 LSB guarantees that the converter will have no missing codes.
DIGITAL INPUT
DIGITAL
y ((DNL)) is the difference between the actual step
p size and
Differential Nonlinearity
the ideal 1 LSB change between two adjacent codes.
DNL error results in:
Smaller or larger step sizes than the ideal
Additive noise/spurs beyond the effects of quantization
Converter Errors (Unipolar)
POSITIVE
GAIN ERROR
NEGATIVE
GAIN ERROR
OUTPUT
VOLTAGE
ACTUAL
IDEAL
DAC Definitions
GAIN AND
OFFSET
ERROR
U
FULL-SCALE
SC
ERROR
Zero-Code Error is typically expressed in LSBs.
DAC Offset Error is a measure of the difference between the actual VOUT
and the ideal VOUT in the linear region of the transfer function. Offset error
can be negative or positive in the DAC and output amplifier.
OUTPUT
VOLTAGE
ACTUAL
IDEAL
ZERO-CODE
ERROR
POSITIVE
OFFSET
DAC CODE
Zero-Code Error is the measured output voltage from VOUT of the DAC
when zero code (all zeros) is loaded to the DAC register.
Offset Error is typically expressed in mV or mA.
mA
DAC Gain Error is a measure of the span error of the DAC. It is the deviation
in slope of the actual DAC transfer characteristic from the ideal.
Gain Error
Ga
o is
s usua
usually
ye
expressed
p essed as a pe
percentage
ce tage o
of tthe
e full-scale
u sca e range.
a ge
Full-Scale Error is a measure of the output error when full-scale code
(0xFFFF) is loaded into the DAC register. Ideally, the output should be VREF
− 1 LSB. (Full-Scale Error = Offset Error + Gain Error)
DAC CODE
Full-Scale Error is typically expressed as a percentage of the full-scale range.
GAIN AND
OFFSET
ERROR
FULL-SCALE
ERROR
OUTPUT
VOLTAGE
ACTUAL
IDEAL
NEGATIVE
OFFSET
AMPLIFIER
FOOTROOM
NEGATIVE
OFFSET
ZERO-CODE
ERROR
Deadband Errors, DACs with integrated output amplifiers will have
of the output amplifier.
amplifier
The number of deadband codes depends on the DAC output voltage span, the
headroom and footroom of the amplifier, and the power supply rails used.
C Definitions
e
to s
ADC Offset Error is the deviation of the first code transition, for example
(000…000) to (000…001) from the ideal (AGND + 1 LSB). Offset error is
typically expressed in LSBs.
DAC CODE
ADC Gain Error is the deviation of the last code transition, for example
(111…110) to (111…111) from the ideal (VREF – 1 LSB) after the offset error is
adjusted out. Gain error for an ADC does not include the reference error
and is typically expressed in LSBs.
www.analog.com/DataConverters
The Essential Guide to Data conversion
Time Domain DAC Output
OVERSHOOT
Clock to output delay
Group delay due to DAC propagation delay
GLITCH
IMPULSE
ENERGY
CLOCK / DATA
FEEDTHROUGH
Settling time
Measured relative to output signal alone
Time between when signal leaves ±0.5 LSB error band to when it
remains within ±0.5 LSB error band of final value
NON-IDEAL
RESPONSE
DNL ERROR
DAC ANALOG OUT
TPUT
SETTLING
ERROR
BAND
IDEAL
RESPONSE
1 LSB
NONLINEAR
SLEWING
Slew rate
Defined as maximum rate of change of voltage or current at output
Specified as V/sec or A/sec depending on DAC output stage
Typically measured for full-scale step size with 10% to 90% error band
CLOCK / DATA
FEEDTHROUGH
CLOCK TO
OUTPUT DELAY
CODE =
ZEROSCALE
CODE =
ZEROSCALE
T1
CODE =
MIDSCALE
CODE =
MIDSCALE
T3
T2
CODE =
MIDSCALE + 1
T4
TIME
T5
Glitch impulse energy
C
Caused
db
by unequall propagation
ti
d
delays
l
within
ithi DAC
Often measured for midscale LSB transition (011..111 to 100..000)
Measured as “area” of glitch impulse with units p/nV-s or p/nA-s
Frequency Domain DAC Output
FULL-SCALE
–x dB FROM FULL-SCALE (dBFS)
–3.9dBc @ FDAC/2
FUNDAMENTAL
SIGNAL
SINC ATTENUATION
DESIRED SIGNAL IMAGES
Sinc(x)
DAC’s time domain step response (zero-order hold) modifies DAC
frequency response
DAC output signals are attenuated by sin(π f/fdac)/(π f/fdac) envelope
Harmonics
Created by DAC’s static and dynamic nonlinearities
AMPL
ITUDE (dB)
Amp
plitude (dB)
SFDR (dBc)
2nd AND 3rd
HARMONICS
DAC CLOCK FEEDTHROUGH
DESIRED SIGNAL
2nd AND 3rd IMAGE HARMONICS
NSD (dBm/Hz)
FDAC/2
FDAC
Frequency
FREQUENCY
g
Images
Duplicate of the desired signal (and its DAC induced harmonics) at
higher Nyquist zones
Images are predicted by sampling theory
SFDR
Measured with single-tone output in first Nyquist band (unit is dBc)
Difference between single-tone
g
amplitude
p
to the next highest
g
spurious
p
tone
Noise Spectral Density (NSD)
Integration
g
of the noise floor in a small frequency
q
y band (unit
(
is
dBm/Hz or nV/rtHz)
Nyquist’s Criteria
AMP ITUDE
AMPLITUDE
NYQUIST ZONE 1
(BASEBAND)
0
Fs
fa IMAGE
FREQUENCY
0.5Fs
NYQUIST ZONE 2
A signal with a maximum frequency fa must be sampled at a rate
Fs > 2fa or information about the signal will be lost because of
aliasing.
aliasing
Aliasing occurs whenever Fs < 2fa.
NYQUIST ZONE 3
The concept
p of aliasing
g is widely
y used in communications
applications such as direct IF-to-digital conversion.
fa
NYQUIST ZONE 4
Undersampled analog signal fa sampled @ Fs has images
0.5,
((aliases)
li
) att | ± KFs ± fa|,
| K=0
5 1
1, 1
5…
1.5
1.5Fs
A signal that has frequency components between fa and fb must
be sampled at a rate Fs > 2 (fb – fa) in order to prevent alias
components from overlapping the signal frequencies.
2Fs
www.analog.com/DataConverters
The Essential Guide to Data conversion
Analog-to-Digital Converter AC Performance Specifications
SNR (Signal-to-Noise Ratio, dB or dBFS)
The ratio of the RMS value of the measured output signal (peak or full
scale) to the RMS sum of all other spectral components excluding the first
6 harmonics and DC.
FULL-SCALE (FS)
( )
INPUT SIGNAL LEVEL (CARRIER)
RMS Signal = (FSR / 2) / √(2), RMS Noise = Qn = q / √(12)
SNR (dB) = RMS Signal / RMS Noise = 20 × log(2(n – 1) × √6)) = 6.02 × n + 1.76
SFDR (dBFS)
SFDR (dBc)
dB
SINAD (Signal-to-Noise Ratio and Distortion, dB)
The ratio of the RMS signal amplitude to the RMS value of the sum of all
other spectral components including harmonics, but excluding DC.
SINAD (dB) = –20 × log (√(10(–SNR W/O DIST/10) + 10(THD/10)))
ENOB (BITS) = (SINAD – 1.76 + 20 × log (FSR/Actual FSR)) / 6.02
WORST SPUR LEVEL
THD (Total Harmonic Distortion, dBc)
The ratio of the RMS sum of the first 6 harmonics to the RMS value of the
measured fundamental.
6TH HAR/20))2 )
2ND HAR/20))2 + (10((–3RD
3RD HAR/20))2 +…
THD (–dB)
( dB) = 20 × log (√((10((–2ND
+ (10((–6TH
SFDR (Spurious-Free Dynamic Range, dB or dBFS)
0, DC
The ratio of the RMS value of the peak signal amplitude (or full-scale) to
the RMS value of the amplitude of the peak spurious spectral component.
The peak spurious component may or may not be a harmonic.
fs
2
FREQUENCY
Oversampling Relaxes Requirements on Baseband Antialiasing Filter
Oversampling Relaxes Requirements on Baseband Antialiasing Filter
fa
A
B
fa
fs – fa
Kfs – fa
Generally an antialiasing filter is required on the analog front end of an ADC.
If the sampling frequency is not much greater than the max input frequency fa, then
the requirements on an antialiasing filter can be severe, as in (A).
g
indicate where the dynamic
y
range
g can be limited by
y signals
g
The dotted regions
outside the bandwidth of interest.
DR
Oversampling relaxes the requirements of the analog antialiasing filter as
shown in (B).
Sigma delta converters are a good example.
Sigma-delta
example
fs
2
fs
STOPBAND ATTENUATION = DR
TRANSITION BAND: fa to fs – fa
CORNER FREQUENCY: fa
Kfs
2
STOPBAND ATTENUATION = DR
TRANSITION BAND: fa to Kfs – fa
Kfs
Outputs of DACs need filtering also, and these are called “anti-imaging”
filters. They serve essentially the same purpose as the antialiasing filter
CORNER FREQUENCY: fa
Theoretical SNR and ENOB Due to Jitter vs. Full-scale Sinewave Analog Input Frequency
Theoretical SNR and ENOB Due to Jitter vs. Full-Scale Sine Wave
A l
Analog
IInputt F
Frequency
130
RMS JITTER
110
0.125 ps
100
0.5 ps
0.25 ps
TOTAL JITTER = t j (RMS)
SNR = 20log 10
1
2 π ft j
The total amount of jitter is dependent on
the effective aperture jitter within the
converter, as well as the external jitter
generated by the sampling clock circuit.
16 BITS
1 ps
90
2 ps
14 BITS
80
12 BITS
70
ENOB
SNR IN dB
S
B
120
dV
TOTAL JITTER =
(ADC APERTURE JITTER) 2 + (SAMPLING CLOCK JITTER) 2
10 BITS
60
ERROR
VOLTAGE
50
40
30
These terms are root sum squared to determine
the total amount of jitter applied to the signal
chain.
1
10
100
1000
ENCODE
dt
www.analog.com/DataConverters
In this example, if a 12-bit ENOB, 74 dB SNR
is desired for the design with an analog
MHz, then the total
input frequency of 100 MHz
jitter required must be 0.5 ps or less.
The Essential Guide to Data conversion
VIN
UNIPOLAR SINGLE-ENDED
FS
0V
UNIPOLAR DIFFERENTIAL
FS
VIN+
VIN+
0V
+
–
−FS/2
VIN−
Differential signals measure the difference in voltage between the
positive and negative input terminals.
The inputs are 180 degrees out of phase with each other.
Many benefits in using differential inputs.
inputs
Vcm
1V
Vcm = 1 VDC
0V
BIPOLAR DIFFERENTIAL
1.5V
VIN+ VIN–
1V p-p
+FS/2
VIN
BIPOLAR SINGLE-ENDED
Single-ended signaling is most common.
p
Example:
p Unipolar
Differential
VIN−
Input transient reduction
Input noise reduction
Signal swing is doubled
1V p-p
VIN–
VIN
+FS/2
0 5V
0.5V
Pseudo differential is a single-ended/differential hybrid.
0V
VIN+
−FS/2
VIN(adc) = VIN+ – VIN– = 2 V p-p
Consider differential common-mode requirements
VIN+
PSEUDO DIFFERENTIAL
Vcm = ((Vp
q
p + Vn)) / 2.
The ADC converts VIN+ – VIN–.
FS
LOW DC
INPUT
VIN−
0V
What Resolution Do I Need? Dynamic Range vs. Signal-to-Noise Ratio Requirements
Dynamic Range (DR) is the difference in level between the highest signal
y the system
y
peak that can be reproduced by
and the amplitude of the
highest spectral component of the noise floor.
DR provides amplitude range so the converter can “see” the signal of interest.
Converter DR is limited by SFDR and, theoretically, by its resolution.
C
Consider
id using
i analog
l
gain
i tto iincrease DR capability
bilit off th
the system.
t
Si
Signal-to-Noise
lt N i R
Ratio
ti (SNR) iis th
t
th
between
the RMS
the diff
difference iin llevell b
signal level and the RMS level of the noise floor, except the first six
harmonics and DC.
0
SNR limits
li it the
th capability
bilit off the
th
converter to see “small” signals.
AAF
DYNAMIC
RANGE
SIGNAL CHAIN
FREQUENCY
Effective Number of Bits
(ENOB) is calculated from SNR:
ENOB = (SNR – 1.76) / 6.02 (bits).
FREQUENCY
74dB = 6.02N + 1.76dB = SNR
40
AM
MPLITUDE (dB)
VGA
DYNAMIC
RANGE
AMPLITUDE
AMPLITUDE
Converter SNR is, theoretically,
limited by its resolution.
Quantization noise of an
SNR = 6.02N + 1.76 (dB),
N = Number of bits.
N = 12 BITS
M = 4096 # OF FFT PTS
FS = 245.76MSPS
FS
BIN SPACING =
4096
20
60
RMS QUANTIZATION NOISE LEVEL
80
74dB
M
33dB = 10log 10 ( 2 ) = FFT NOISE FLOOR (PER BIN)
100
107dB
120
140
Example: 10-bit ADC with an FSR = 4 V p-p has an LSB = 3.9 mV p-p or 4/2BITS.
Therefore, 4 V / 3.9 mV = 1024 codes. This can also be expressed in dB or 20 × log
(1024) = 60 dB.
FS
2
78 = 10log10(FS/2) = NOISE FLOOR (PER Hz)
152dB
160
Quantization:
The Size of a Least Significant Bit (LSB)
FREQUENCY (Hz)
Converter Circuits
DAC Amplifier Coupled Circuit
500Ω
VDD
AVDD
RESOLUTION N
2N
VOLTAGE
(2 V/10 V FS)
ppm FS
% FS
dBFS
2-bit
4
0.5/2.5 V
250,000
25
–12
4-bit
16
125/625 mV
62,500
6.25
–24
6-bit
64
31.3/156 mV
15,625
1.56
–36
8-bit
8
bit
256
7.8/39.1 mV
3906
0.39
–48
48
10-bit
1024
2/9.77 mV
977
0.098
–60
0.49/2.44 mV
244
0.024
–72
12-bit
4096
14-bit
16,384
122/610 µV
61
0.0061
16-bit
65,536
30.5/153 µV
15
0.0015
–96
18-bit
262,144
7.6/38 µV
4
0.0004
–108
20-bit
1,048,576
1.9/9.54 µV
1
0.0001
–120
22-bit
4,194,304
0.47/2.38 µV
0.24
0.000024
–132
24-bit
16,777,216
119/596 nV*
0.06
0.000006
–144
*600 nV is the Johnson Noise in a 10 kHz BW of a 2.2 kΩ resistor @ 25°C.
–84
225Ω
Ω
IOUTB
DAC
AMP
225Ω
IOUTA
GND
25Ω
COPT
AVSS
25Ω
500Ω
ROPT
AVDD
1kΩ
AMP-AVDD
AVDD
205Ω
24Ω
ANALOG
INPUT
200Ω
62Ω 10kΩ
0.1µF
+VS
VOCM
10kΩ
200Ω
G = UNITY
–VS
27Ω
Remember: 10 bits and 10 V FS yields an LSB of 10 mV, 1000 ppm, or 0.1%.
(All other values may be calculated by powers of 2.)
10kΩ
REQUIRED FOR
10kΩ
24Ω
205Ω
0.1µF
1 F
0.1µF
0
33Ω
VIN+
2p
DRVDD
BUFFERED OR