Mixed-Signal-Electronics
PD Dr.-Ing. Stephan Henzler
Stephan Henzler Mixed-Signal-Electronics 2011/12 1
Chapter 4
Data Converter Fundamentals
"He who loves practice without theory is like the
sailor who boards ship without a rudder and
compass and never knows where he may cast“
Leonardo Davinci
Stephan Henzler Mixed-Signal-Electronics 2011/12 2
Ideal Digital-to-Analog Conversion
Stephan Henzler Mixed-Signal-Electronics 2011/12 3
DAC Offset Error
Stephan Henzler Mixed-Signal-Electronics 2011/12 4
DAC Gain Error
Stephan Henzler Mixed-Signal-Electronics 2011/12 5
Digital-to-Analog Converter Model
ideal model 1st order model
Gain of ideal converter depends on interpretation of DAC
input bits. Here, a fractional interpretation is used.
Stephan Henzler Mixed-Signal-Electronics 2011/12 6
Ideal Analog-to-Digital Conversion
Stephan Henzler Mixed-Signal-Electronics 2011/12 7
Analog-to-Digital Converter Model I
Depending on implementation quantization block can be
– round function
– floor function
– ceil function
Stephan Henzler Mixed-Signal-Electronics 2011/12 8
Quantization in A/D-Converters
Stephan Henzler Mixed-Signal-Electronics 2011/12 9
Quantization in A/D-Converters
Consider one saw tooth, e.g.
Mean quantization error:
Power of quantization error:
Stephan Henzler Mixed-Signal-Electronics 2011/12 10
Quantization in A/D-Converters
Signal-to-(Quantization)-Noise Ratio S(Q)NR:
Stephan Henzler Mixed-Signal-Electronics 2011/12 11
ADC Offset Error
Stephan Henzler Mixed-Signal-Electronics 2011/12 12
ADC Gain Error
Stephan Henzler Mixed-Signal-Electronics 2011/12 13
Analog-to-Digital Converter Model II
Stephan Henzler Mixed-Signal-Electronics 2011/12 14
Non-Linearity in Data Converters
Stephan Henzler Mixed-Signal-Electronics 2011/12 15
Example of Non-Linear Transfer Characteristic
x non-linear y
system
Stephan Henzler Mixed-Signal-Electronics 2011/12 16
Nonlinearity in Data-Converters
Accuracy:
The absolute inaccuracy is
defined as the maximum
deviation of the actual analog
value from the ideal one. It
includes offset, gain and
linearity errors and may be
different for every individual
quantization value.
The relative inaccuracy is
defined as the deviation that
remains after offset and gain
error have been removed.
Stephan Henzler Mixed-Signal-Electronics 2011/12 17
DAC Nonlinearity: Integral Nonlinearity
Integral Nonlinearity (INL) Error:
The INL error is defined as the deviation of each analog value from a straight
line. If the straight line through the endpoints of the converter’s transfer curve
is chosen (left) as reference, the INL equals the relative inaccuracy.
Alternatively a regression line can be used as reference (right). It depends on
the application which definition should be applied.
Stephan Henzler Mixed-Signal-Electronics 2011/12 18
DAC Nonlinearity: Differential Nonlinearity
Differential Nonlinearity (DNL):
Deviation of the analog step size
from 1VLSB (after removal of gain
and offset error). Like the INL the
DNL value is defined individually
for each digital word
Monotonicity:
The output signal of a monotonic
D/A converter increases or
remains at least unchanged as
the input signal increases.
Stephan Henzler Mixed-Signal-Electronics 2011/12 19
ADC Nonlinearity: Inaccuracy
Accuracy:
The absolute inaccuracy is
defined as the maximum
deviation of the actual analog
value from the ideal one. It
includes offset, gain and
linearity errors and may be
different for every individual
quantization value.
The relative inaccuracy is
defined as the deviation that
remains after offset and gain
error have been removed.
Stephan Henzler Mixed-Signal-Electronics 2011/12 20
ADC Nonlinearity: Integral Nonlinearity
Integral Nonlinearity (INL):
The INL error is defined as the
deviation of each analog value
from a straight line (after
removal of gain and offset error).
If the straight line through the
endpoints of the converter’s
transfer curve is chosen as
reference, the INL equals the
relative inaccuracy. Alternatively
a regression line can be used as
reference. It depends on the
application which definition
should be applied.
Stephan Henzler Mixed-Signal-Electronics 2011/12 21
ADC Nonlinearity: Differential Nonlinearity
Differential Nonlinearity (DNL):
Deviation of the analog step size
from 1VLSB (after removal of gain
and offset error). Like the INL the
DNL value is defined individually
for each digital word
Missing Codes:
If one step vanishes completely
the ADC is said to have a missing
code. In practice the criterion for a
missing code is a step width
smaller than 0.1 x VLSB
Stephan Henzler Mixed-Signal-Electronics 2011/12 22
Dynamic ADC Measurement
ADC converter characteristics
describe static measurements
In dynamic measurements a input
voltage waveform (usually a sine signal)
is converted continuously and
analyzed in the frequency domain.
(much more realistic, noise, crosstalk, etc.)
Stephan Henzler Mixed-Signal-Electronics 2011/12 23
Interpretation of the Output spectrum
test signal Robertson, ISSCC, 2002
A sin(wt)
[FS]
harmonics 0.5 x
sampling
frequency
noise floor
Stephan Henzler Mixed-Signal-Electronics 2011/12 24
Harmonic Distortion Caused by Nonlinearity
Nonlinearity causes harmonic distortion
Spectrum allows computation of
– signal power
– power of harmonic components
– noise power
Signal-to-Noise Ratio:
Psignal
SNR 10 log10
Pnoise
Signal-to-Noise-and-Distortion Ratio
Psignal
SNDR 10 log10
Pnoise Pharmonics
Stephan Henzler Mixed-Signal-Electronics 2011/12 25
Harmonic Distortion Caused by Nonlinearity
Total Harmonic Distortion (THD)
V 2 V 2 V 2
THD 20 log10 2 3 4
V1
Second Harmonic Distortion (SHD)
V2
SHD 20 log10
V1
Stephan Henzler Mixed-Signal-Electronics 2011/12 26
Effective Number of Bits
An ideal ADC suffers only from quantization noise
SNR SQNR 6.02dB N 1.76dB
Actual ADCs have lots of impairments (noise, nonlinearity)
that limit the effective resolution.
Effective number of Bits:
SNDR 1.76dB
ENOB
6.02dB
Note: The number of bits coming out of an ADC doesn‘t say
anything about the actual resolution
Stephan Henzler Mixed-Signal-Electronics 2011/12 27
Signal-to-Noise Ratio over Input Signal
Stephan Henzler Mixed-Signal-Electronics 2011/12 28
Other Useful Figures in Mixed-Signal
Crest Factor = Peak-to-Average Ratio
Measure for dynamics of a signal defined by the quotient of
the peak amplitude and the RMS value
max x(t )
C
2
x (t )
ADC Figure-of-Merit:
Measure for the efficiency of an ADC
power
FOM
frequency 2 ENOB
Stephan Henzler Mixed-Signal-Electronics 2011/12 29
Why is Linearity so Important?
Let‘s consider a basic example of third order non-linearity
x non-linear y
system
The following equalities are useful
Harmonic Distortion:
A single tone transforms into a dc offset and multiple tones at
multiples of the original input frequency
Stephan Henzler Mixed-Signal-Electronics 2011/12 30
Non-Linearity Gain Compression
Consider only the fundamental (i.e. for low amplitude)
For small amplitude A the gain of the system is a1
With increasing amplitude A, the high order term becomes
noticeable, i.e. the gain becomes
For a compressive system
(a3<0) the gain for the
fundamental decreases with
increasing amplitude and
may even vanish completely
Stephan Henzler Mixed-Signal-Electronics 2011/12 31
Superposition of two Frequencies
Stephan Henzler Mixed-Signal-Electronics 2011/12 32
Superposition of Two Frequencies
Stephan Henzler Mixed-Signal-Electronics 2011/12 33
Non-Linearity Desensitization
Consider superposition of weak signal and a strong interferer
Strong signal causes large excitation around operating point
effective gain reduction
Small signal gain may become even zero
blocking
Stephan Henzler Mixed-Signal-Electronics 2011/12 34
Non-Linearity Cross Modulation
Consider again superposition of weak signal and a strong
interferer, but now with modulation of interferer
Amplitude variation m2 may be either noise or modulation
Signal in band around 1
Modulation/noise of the carrier 2 is transferred 1
Effect occurs if multiple channels are processed by same
non-linear system, e.g. non-linear amplifier in wireless base
station
Stephan Henzler Mixed-Signal-Electronics 2011/12 35
Non-Linearity Intermodulation
Harmonics can be used to quantify the non-linearity of mixed-
signal system.
If the system is frequency selective, e.g. a low-pass filter
some or even all harmonics are damped
system seems to be less non-linear than it actually is.
Measure non-linearity in band of interest, i.e. without
damping effects Intermodulation Experiment
Two-Tone Test:
Stephan Henzler Mixed-Signal-Electronics 2011/12 36
Non-Linearity Intermodulation
Two tones passing a non-linear system result in
– Harmonics of first frequency
– Harmonics of second frequency
– Intermodulation products (caused by mixing signals)
If the frequency difference is small, i.e. 2 - 1 << 1
the third order intermodulation frequencies lie nearby
– lower 3rd order IM product: 21 – 2 = 1 – (2 - 1)
– upper 3rd order IM product: 22 - 1 = 2 + (2 - 1)
– negligible impact of frequency dependent transfer function
Stephan Henzler Mixed-Signal-Electronics 2011/12 37
Non-Linearity Intermodulation
Intermodulation distortion = 3rd IM product / fundamental
(Amplitude must be known for interpretation)
Assume small amplitudes, i.e. the amplitude of the
fundamental can be expressed by a1A and the IM3 is 3/4a3A3
Third order intercept point describes the amplitude where the
IM3 equals the fundamental
(Hypothetical as for high amplitudes higher order non-linearity becomes noticeable, i.e. meas. is done for small amplitudes followed by extrapolation)
Stephan Henzler Mixed-Signal-Electronics 2011/12 38
Non-Linearity: Further Reading
Recommended Literature:
– Razavi. RF Microelectronics. Prentice Hall.
– Robertson. Specification and Figures of Merit for
Mixed-Signal Circuits. ISSCC Tutorial 2002.
– Razavi. Principles of Data Conversion System Design.
Wiley, 1994.
Stephan Henzler Mixed-Signal-Electronics 2011/12 39
Binary Number Representation
# # Sign 1‘s 2‘s Offset
normalized magnitude complement complement binary
+7 +7/8 0111 0111 0111 1111
+6 +6/8 0110 0110 0110 1110
+5 +5/8 0101 0101 0101 1101
+4 +4/8 0100 0100 0100 1100
+3 +3/8 0011 0011 0011 1011
+2 +2/8 0010 0010 0010 1010
+1 +1/8 0001 0001 0001 1001
+0 0 0000 0000 0000 1000
-0 0 1000 1111
-1 -1/8 1001 1110 1111 0111
-2 -2/8 1010 1101 1110 0110
-3 -3/8 1011 1100 1101 0101
-4 -4/8 1100 1011 1100 0100
-5 -5/8 1101 1010 1011 0011
-6 -6/8 1110 1001 1010 0010
-7 -7/8 1111 1000 1001 0001
-8 -8/8 1000 0000
Stephan Henzler Mixed-Signal-Electronics 2011/12 40
